Choice Over Control: How Users Write with Large Language Models using Diegetic and Non-Diegetic Prompting
Open accessAbstract
We propose a conceptual perspective on prompts for Large Language Models (LLMs) that distinguishes between (1) diegetic prompts (part of the narrative, e.g. “Once upon a time, I saw a fox...”), and (2) non-diegetic prompts (external, e.g. “Write about the adventures of the fox.”). With this lens, we study how 129 crowd workers on Prolific write short texts with different user interfaces (1 vs 3 suggestions, with/out non-diegetic prompts; implemented with GPT-3): When the interface offered multiple suggestions and provided an option for non-diegetic prompting, participants preferred choosing from multiple suggestions over controlling them via non-diegetic prompts. When participants provided non-diegetic prompts it was to ask for inspiration, topics or facts. Single suggestions in particular were guided both with diegetic and non-diegetic information. This work informs human-AI interaction with generative models by revealing that (1) writing non-diegetic prompts requires effort, (2) people combine diegetic and non-diegetic prompting, and (3) they use their draft (i.e. diegetic information) and suggestion timing to strategically guide LLMs.
Verified 2026-10-05 · 1 reference not checked yet
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
0 of 57 cited sources checked · 0 supported · 0 flagged
- Findingown data
An LMM fitted on task completion time for the suggestion UIs did not reveal significant effects.
“an LMM fitted on this data for the suggestion UIs did not reveal significant effects (Table 1, row 2).”
Quote in the paperNot a substantial claim - Findingown data
The grand mean acceptance rate across UIs was 0.64 (SD 0.29).
“The grand mean acceptance rate was 0.64 (SD 0.29).”
Quote in the paperNot a substantial claim - Findingown data
Showing one suggestion instead of three significantly decreased the chance of acceptance.
“showing one suggestion (instead of three) significantly decreased the chance of acceptance”
Quote in the paperSubstantial claim - Findingnone given
Allowing users to provide non-diegetic prompts reduces the acceptance-rate gap between single and multiple suggestions by increasing acceptance for the single-suggestion UI.
“yet allowing them to provide non-diegetic prompts reduces the gap in acceptance rates by boosting it for the UI with single suggestions.”
Quote in the paperSubstantial claim - Findingown data
Writing with three suggestions took significantly longer than writing without suggestions.
“Here we found that writing with three suggestions took significantly longer than without suggestions.”
Quote in the paperNot a substantial claim
- Findingown data
Single suggestions were rated as significantly less helpful than three suggestions (x=0.60, p=0.02).
“less helpful (x=0.60,p=0.02)”
Quote in the paperSubstantial claim - Findingown data
Writing with multiple suggestions leads to significantly longer texts.
“writing with multiple suggestions leads to significantly longer texts.”
Quote in the paperSubstantial claim - Findingreasoning
Users generally prefer selecting from multiple suggestions rather than controlling the system through non-diegetic prompts.
“Users overall prefer choosing from multiple suggestions over controlling them via non-diegetic prompts.”
Quote in the paperSubstantial claim - Findingnone given
Users prefer a UI with multiple suggestions over one with a single suggestion.
“Users also prefer the UI with multiple suggestions over seeing single ones”
Quote in the paperSubstantial claim - Findingown data
Single suggestions were rated as significantly more distracting than three suggestions (x=1.95, p<0.005).
“Single suggestions were rated as significantly more distracting (x=1.95,p<0.005)”
Quote in the paperSubstantial claim - Findingown data
The mean acceptance rate for suggestion requests with a written instruction was 0.64 (SD 0.33).
“The mean for suggestion requests with a written instruction was in line with this (0.64, SD: 0.33).”
Quote in the paperNot a substantial claim - Findingnone given
Users employ non-diegetic prompts to request inspiration, topics, or facts from the LLM.
“They use non-diegetic prompts to ask the LLM for inspiration, topics or facts.”
Quote in the paperSubstantial claim - Findingown data
No significant differences in text length were found between the suggestion UIs.
“Empty cells indicate no significant results.”
Quote in the paperNot a substantial claim - Findingown data
Giving instructions was rated significantly better on being able to influence the suggested text (x=1.92, p<0.001).
“giving instructions was rated significantly better on being able to influence the suggested text (x=1.92,p<0.001).”
Quote in the paperSubstantial claim - Findingown data
Six participants described a trial-and-error approach to find out how to best write instructions.
“Six participants described a trial-and-error approach to find out how to best write instructions.”
Quote in the paperNot a substantial claim - Findingown data
For suggestions based on user instructions, the acceptance rate in this study (64%) was higher than in a related study design where users could enter requests in a sidebar (17.6%).
“For suggestions based on user instructions, our rate (64 %) is higher than in a related study design where users could enter requests in a sidebar (17.6 % in [48])”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [48]
Not checked in the free run.
- Findingown data
Users saw the AI as a transcriber that improved input efficiency.
“they perceived the AI as a transcriber for input efficiency (Section 8).”
Quote in the paperNot a substantial claim - Findingown data
The pre-study indicated that such examples are necessary to help people begin using the feature.
“Our pre-study showed that such examples are needed to help people get started with this new feature.”
Quote in the paperNot a substantial claim - Findingown data
For the two questions focused on instructions (Q15 and Q16), no significant differences were found between the single-suggestion and three-suggestion UIs.
“For these two questions, we found no significant differences between 𝑠 1 and 𝑠 3 .”
Quote in the paperNot a substantial claim - Findingown data
No significant differences in task completion times were found between the four UIs with suggestions.
“No sig.differences in task completion times were found between the four UIs with suggestions.”
Quote in the paperNot a substantial claim - Findingown data
Twenty-five participants said they did not use instructions much because, for example, they were not good at thinking of them.
“Fittingly, 25 participants said they did not use instructions much because, for example,”
Quote in the paperNot a substantial claim - Findingown data
Single suggestions were rated as feeling significantly less in control than three suggestions (x=0.67, p=0.03).
“feeling less in control (x=0.67,p=0.03)”
Quote in the paperSubstantial claim - Findingown data
In the collected non-diegetic prompts, users asked the AI to generate inspirational ideas.
“non-diegetic prompts from our dataset show that users requested the AI to propose inspirational ideas.”
Quote in the paperNot a substantial claim - Findingown data
Users wrote about 1.5 more words in the sentence before requesting single suggestions compared to multiple suggestions.
“Users wrote about 1.5 more words in the sentence before requesting single suggestions, compared to multiple ones.”
Quote in the paperSubstantial claim - Findingown data
UIs that allowed users to enter instructions received ratings of significantly more manual editing of suggestions (x=1.54, p=0.02).
“the UIs that allowed users to enter instructions to the AI received ratings of manually editing suggestions significantly more (x=1.54,p=0.02)”
Quote in the paperSubstantial claim - Findingown data
Single suggestions were rated as providing significantly less diverse suggestions than three suggestions (x=0.67, p=0.04).
“and providing less diverse suggestions (x=0.67,p=0.04).”
Quote in the paperNot a substantial claim - Findingown data
Some participants said writing with instructions felt less natural (coded 3 times).
“Some said writing with instructions felt less natural (coded 3 times)”
Quote in the paperNot a substantial claim - Findingown data
The majority of participants preferred multiple suggestions (75 people) over single suggestions (23 people).
“The majority preferred multiple suggestions (75 people stated this preference vs 23 for single suggestions).”
Quote in the paperNot a substantial claim - Findingown data
Writing with three suggestions took longer than writing without suggestions.
“Writing with three suggestions took longer than without suggestions.”
Quote in the paperSubstantial claim - Findingown data
Users sometimes gave only partial phrases or keywords, or asked for word choices, which placed the AI in the role of turning abstract ideas into full sentences.
“Sometimes users also only provided partial phrases or keywords, or asked for word choices, which puts the AI into the role of translating these abstract ideas into full sentences.”
Quote in the paperNot a substantial claim - Findingown data
Overall, suggestions received high ratings.
“Nevertheless, suggestions were rated highly overall (Section 7.5).”
Quote in the paperNot a substantial claim - Findingown data
Single suggestions were rated as leading to significantly more manual editing than three suggestions (x=1.91, p<0.005).
“leading to more manual editing (x=1.91,p<0.005)”
Quote in the paperSubstantial claim - Findingown data
For argumentative writing and when people already had an opinion about a topic, they felt the AI was distracting if it proposed something other than what they had in mind.
“For argumentative writing and, more generally, when people already had an opinion about a topic, they felt that the AI was distracting if it proposed something other than what participants had in mind.”
Quote in the paperSubstantial claim - Findingown data
Thirty-three participants touched on challenges of learnability and writing instructions.
“Several participants (33) touched on challenges of learnability and writing instructions”
Quote in the paperNot a substantial claim - Findingnone given
For all suggestion UIs, the majority did not find them distracting.
“That said, note that for all suggestion UIs, the majority did not find them distracting.”
Quote in the paperNot a substantial claim - Findingcitation
Participants accepted suggestions at a rate comparable to related work: 74% in this study versus 72% in a cited study.
“participants accepted suggestions at a rate comparable with related work (with multiple suggestions and explicit request via tab key: 74 % here vs 72 % in [26]).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [26]
Not checked in the free run.
- Findingown data
Participants went beyond the provided examples in what they tried.
“Nevertheless, people experimented beyond these examples (e.g.questions, writing help, advice, etc.; see Table 2).”
Quote in the paperNot a substantial claim - Findingown data
UIs that allowed users to enter instructions were rated as significantly more distracting (x=2.03, p<0.0005).
“and being significantly more distracting (x=2.03,p<0.0005).”
Quote in the paperSubstantial claim - Findingown data
Users also deliberately consider and shape their own text to steer LLMs through diegetic information.
“our findings reveal that users additionally think about and shape their text to guide LLMs through diegetic information.”
Quote in the paperSubstantial claim - Findingnone given
Writing instructions to the AI demands effort, which includes switching between diegetic and non-diegetic writing.
“First, writing instructions to the AI requires effort, including switching between diegetic and non-diegetic writing.”
Quote in the paperSubstantial claim - Findingown data
Showing one suggestion rather than three lowered the probability of requesting suggestions at the start of a new sentence.
“Showing one suggestion (instead of three) decreased the chance of requesting suggestions at the beginning of a new sentence.”
- Also stated in: § 10 CONCLUSION · plain
“Single suggestions were requested less often at the start of sentences (bottom left), about 1.5 words later in a sentence (bottom center), and less often after transition words (bottom right).”
Quote in the paperSubstantial claim - Also stated in: § 10 CONCLUSION · plain
- Findingown data
With a UI supporting both prompt types and using GPT-3, participants favored selecting among multiple suggestions over writing instructions.
“With our UI design that allows for both types, using GPT-3, participants preferred choosing from multiple suggestions over writing instructions.”
Quote in the paperSubstantial claim - Findingown data
People employ two kinds of prompting to steer LLMs in generating text.
“people use two types of prompting to guide LLMs for text generation.”
Quote in the paperSubstantial claim - Findingown data
Showing one suggestion rather than three lowered the probability of requesting suggestions after a transition word.
“Showing one suggestion (instead of three) decreased the chance of requesting suggestions after a transition word.”
Quote in the paperSubstantial claim - Findingown data
Writing time did not differ much across the conditions.
“Writing time did not vary much and texts were slightly shorter with single suggestions and instructions (top center/right).”
Quote in the paperNot a substantial claim - Findingown data
Writing with multiple suggestions produced slightly longer texts.
“However, writing with multiple suggestions leads to slightly longer texts.”
Quote in the paperNot a substantial claim - Findingown data
The keyword style of prompt was found to be more common than the imperative style.
“PEER is trained on the imperative-style but we found the keyword-style to be more common.”
Quote in the paperSubstantial claim - Findingown data
Showing one suggestion rather than three raised the number of words into a sentence after which suggestions were requested.
“Showing one suggestion (instead of three) increased the number of words in a sentence after which suggestions were requested.”
Quote in the paperSubstantial claim - Findingown data
Writers strategically guide LLMs using their draft and the timing of their suggestion requests.
“Third, writers use their draft (i.e.diegetic information) and suggestion timing to strategically guide LLMs, based on our analysis of when people request suggestions, as well as their self-reflection in comments.”
Quote in the paperSubstantial claim - Premisecitation
Writing interfaces that allow users to explicitly enter prompts often use a mix of diegetic and non-diegetic information.
“In contrast, writing interfaces that indeed allow users to explicitly enter prompts often use a mix of diegetic and non-diegetic information.”
Quote in the paperNot a substantial claim - Premisecitation
Systems whose UI did not explicitly afford text prompts implicitly used only diegetic information as input to the LLM.
“Therefore, systems in which the UI did not afford text prompts explicitly made the implicit choice of only using diegetic information as their input to the LLM [6,7,26,41].”
Quote in the paperNot a substantial claim0 of 4 sources checked - Premisecitation
Earlier work mainly focused on single-word suggestions.
“Earlier work mainly focused on single word suggestions [11,12,18,34].”
Quote in the paperNot a substantial claim0 of 4 sources checked - Premisecitation
Traditionally, writing systems mainly use diegetic information, predicting text based only on the preceding text.
“Traditionally, systems mainly use diegetic information, that is, they predict text based (only) on the preceding text [11,12,18,34].”
Quote in the paperNot a substantial claim0 of 4 sources checked - Interpretationreasoning
Writers use diegetic information to guide LLMs.
“Together with people's comments, this indicates that writers consider diegetic information to guide LLMs.”
Quote in the paperSubstantial claim - Interpretationreasoning
Those who saw the system as a transcriber were more critical because it would only be good if it is fast and predicts exactly what they want.
“The former are more critical about the system since it would only be good if it is fast and predicts exactly what they want.”
Quote in the paperNot a substantial claim - Interpretationown data
It is feasible to build and closely study interactive applications on top of existing models.
“our work shows that it is possible to construct and study in detail interactive applications built on existing models.”
Quote in the paperSubstantial claim - Interpretationown data
People specify more diegetic information to offset the lack of suggestion choice in UIs that display only a single suggestion.
“Our results add to the literature on writing with AI by revealing that people specify more diegetic information to offset the lack of suggestion choice in UIs that display only a single suggestion.”
Quote in the paperSubstantial claim - Interpretationown data
Switching from diegetic writing to non-diegetic instructing forces writers to shift from thinking about their narrative or argument to thinking about instructions to the system.
“Crucially, switching from diegetic writing to non-diegetic instructing forces writers to shift from thinking about their narrative or argument to thinking about instructions to the system.”
Quote in the paperSubstantial claim - Interpretationown data
Non-diegetic prompts can move how users perceive the AI's role toward that of a writing collaborator.
“Thus, non-diegetic prompts may shift perception of the AI's role towards a writing collaborator.”
Quote in the paperNot a substantial claim - Premisecitation
Transcription tasks have historically been the dominant paradigm in text entry research.
“Historically, transcription tasks have dominated text entry research (cf.[45]).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [45]
Not checked in the free run.
- Interpretationown data
Participants overall preferred choosing from multiple text suggestions over actively writing instructions in short creative and argumentative writing tasks.
“Participants overall preferred choosing from multiple text suggestions presented to them, over actively writing instructions, in short creative and argumentative writing tasks.”
Quote in the paperSubstantial claim- Rests on an assumption: Suggestion acceptance rate is a valid indicator of users' preference for a suggestion UI.
- Rests on an assumption: The writing tasks used in the study are representative of short creative and argumentative writing tasks in general.
- Interpretationown data
Times and texts, in combination with the comments, support the conclusion that participants took the tasks seriously.
“Moreover, times and texts, in combination with the comments, further support the conclusion that participants took the tasks seriously (see Section 7.7).”
Quote in the paperNot a substantial claim - Interpretationown data
Writing effective non-diegetic prompts is difficult.
“Even after making that shift, then writing effective non-diegetic prompts is difficult, adding to related findings in the literature [48]:”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [48]
Not checked in the free run.
- Premisecitation
Schick et al.'s LLM PEER is explicitly trained to follow non-diegetic prompts for text revision.
“In recent work by Schick et al. [38], their LLM”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [38] source
Not checked in the free run.
- Classified as probably not a substantial claim. 54
- May combine two claims. 37
- The quote only partly states this claim. 8
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingown dataParticipants followed the task description of writing for five minutes, and writing with the suggestion UIs took slightly longer.
“Participants followed the task description of writing for five minutes, and writing with the suggestion UIs took slightly longer.”
Quote in the paperNot a substantial claimThe quote only partly states this claim.
Findingown dataSingle suggestions are requested less often after transition words than multiple suggestions.
“and less frequently at transition words”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataThe grand mean number of words per text was 134 (SD 54).
“The grand mean number of words per text was 134 words (SD 54).”
Quote in the paperNot a substantial claimMay combine two claims.
Findingnone givenNon-diegetic prompts increase the effort required to learn how to formulate them and to switch between diegetic and non-diegetic writing.
“Non-diegetic prompts increase effort, for learning how to formulate them and switching between diegetic and non-diegetic writing.”
Quote in the paperNot a substantial claimThe quote only partly states this claim.
Findingnone givenSingle suggestions are requested later in sentences than multiple suggestions.
“single suggestions are triggered later in sentences”
Quote in the paperNot a substantial claimThe quote only partly states this claim.
Findingown dataSingle suggestions are requested less often at the start of sentences than multiple suggestions.
“and to start sentences”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataTexts are significantly shorter when writing with single suggestions or with a UI allowing for instructions.
“texts are significantly shorter when writing with single suggestions or with a UI allowing for instructions.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataSuggestions received favourable ratings by the majority and had almost no perceived grammatical or factual errors.
“Descriptively, suggestions received favourable ratings by the majority and had almost no perceived grammatical or factual errors.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataSubmitted texts contained 87,640 words in total, including text from accepted suggestions.
“In total, submitted texts contained 87,640 words, including text from accepted suggestions.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataAmong the texts that did not align with selected topics, 13 out of 23 were written for the 'shapeshifter' category.
“Of these, the majority (13 out of 23) were written for the category of”
Quote in the paperNot a substantial claimMay combine two claims. The quote only partly states this claim.
Findingown dataThe main reasons for using instructions were getting more suitable suggestions (coded 12 times), inspiration for words (coded 29 times), and delegating tasks such as coming up with places, names, or synonyms (coded 5 times).
“inspiration for words (coded 29 times)”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataOther instruction cases included multiple words used to describe something.
“Other cases included instructions consisting of multiple words to describe something”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataNine participants explicitly reflected on distraction.
“Nine participants explicitly reflected on distraction.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataIn the study, showing one suggestion instead of three decreased the chance of suggestion acceptance, and this decrease was larger without instructions than with instructions.
“Showing one suggestion (instead of three) decreases chance of acceptance; more so without instructions than with them.”
- Also stated in: § Section Aspect · plain
“Suggestion acceptance 𝑠 1 (𝛽=-1.26,SE=0.11,CI 95% =[-1.48,-1.03], p<.0001) Number * Instruction (𝛽=0.72,SE=0.17,CI 95% =[0.39, 1.05], p<.0001)”
- Also stated in: § Accepted text suggestions · plain
“Instructions increased acceptance rates for single suggestions -although these still did not reach the rate for multiple suggestions (Section 7.1).”
Quote in the paperSubstantial claim- Also stated in: § Section Aspect · plain
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.
Findingown dataParticipants sent 397 instructions to the system in total, with an average of 3.08 instructions per person (SD 3.40).
“In total, participants used the non-diegetic prompting option to send 397 instructions to the system, with an average of 3.08 instructions per person (SD: 3.40).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataThe most common instruction style was single keywords or comma-separated lists of keywords, with 171 usages.
“The most common one (171 usages) was to use single keywords (or comma-separated lists of keywords).”
Quote in the paperNot a substantial claimMay combine two claims.
FindingreasoningSingle suggestions were requested less frequently to start a new sentence and to continue after a transition word.
“Moreover, single suggestions were requested less frequently to start a new sentece and to continue after a transition word.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataParticipants mentioned inspiration (coded 39 times), overcoming writer's block (coded 3 times), and finishing sentences (coded 7 times) as reasons for using suggestions.
“Participants commented on why and how to use suggestions, mentioning inspiration (coded 39 times), overcoming writer's block (coded 3 times), or finishing sentences (coded 7 times)”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingown dataComing up with instructions can disrupt the writing flow (coded 3 times) and thus reduce efficiency or not be worth the effort.
“It was also reported that coming up with instructions can disrupt the writing flow (coded 3 times) and thus reduces efficiency, or is not worth the effort.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataUsers asked the AI for an opinion or advice, or to look up information.
“people asked the AI for an opinon or advice, or to lookup information.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFourteen participants reflected on benefits for both multiple and single suggestions.
“Fourteen participants reflected on benefits for both”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataApproximately 3.5% of all texts (23 out of 645) did not align with the selected topics.
“Approximately 3.5% of all texts (23 out of 645) did not align with the selected topics.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingnone givenSuggestions were at times repetitive, such as similar items within one list, or repeated the instruction text.
“Sometimes, suggestions were repetive (e.g.similar ones in one list) or repeated the instruction text (which seems unhelpful).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataEleven participants commented on control and influencing suggestions.
“Eleven participants commented on control and influencing suggestions.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataOn average, participants entered an instruction text in 0.19 of suggestion requests (SD 0.17) for tasks that offered instructions, meaning about every fifth request used instructions.
“On average, participants had a ratio of 0.19 (SD: 0.17) of entering an instruction text when requesting suggestions, for those tasks that offered to do so.”
Quote in the paperNot a substantial claimMay combine two claims. The quote only partly states this claim.
Findingown dataEighteen participants explicitly commented on suggestion quality: eight were negative, nine felt suggestions were hit or miss, and two left positive comments.
“Eighteen participants explicitly commented on suggestion quality: Eight were negative”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataMany participants struggled with writing non-diegetic prompts and recognized that they did so in self-reflection.
“Many participants struggled with this and recognised that they did so in self-reflection (Section 8.2, 8.3, 8.4).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.
Findingown dataThe comments indicate two fundamental views on the role of the AI: some expected the system to serve efficiency, while others saw it as serving inspiration.
“The comments indicate two fundamental views on the role of the AI: Some expected the system to serve efficiency.”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataThe main reasons for preferring multiple suggestions were higher chances of finding fitting suggestions (coded 36 times) and more inspiration (coded 8 times).
“Main reasons were higher chances of finding fitting suggestions (coded 36 times) and more inspiration (coded 8 times).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
FindingcitationThe mean number of spelling and grammar mistakes per word was 0.0025, comparable to values reported in previous research.
“The mean number of spelling and grammar mistakes per word was 0.0025, which is comparable to values reported in previous research [26].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [26]
Not checked in the free run.
May combine two claims.
Findingown dataSome people saw the system as serving input efficiency (a transcriber), while others saw it as providing inspiration (a proposer).
“Some people clearly saw the system as something that serves input efficiency (i.e.transcriber), whereas others saw it as providing inspiration (i.e.proposer).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingown dataThe mean instruction length was 14.20 characters (SD 9.01) and 2.52 words (SD 1.74).
“The mean instruction length was 14.20 characters (SD: 9.01) and 2.52 words (SD: 1.74).”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataThe non-diegetic prompts collected in the study show that users experimented with different styles.
“More positively, the non-diegetic prompts collected in our study show how users experimented with different styles.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataAn imperative instruction style occurred 59 times.
“We also found an imperative style with 59 occurrences”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataParticipants formulated a question in 12 instruction cases.
“In 12 cases, participants formulated a question”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataParticipants who preferred single suggestions found them more intuitive (coded 7 times), faster to work with (coded 3 times), or less distracting (coded 3 times).
“Those preferring single suggestions found them more intuitive (coded 7 times), faster to work with (coded 3 times) or less distracting (coded 3 times)”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataThe study also yielded rich qualitative feedback.
“Nevertheless, we received rich qualitative feedback as well (Section 8).”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataOffering users the option to write instructions raises the acceptance rate of suggestions for single suggestions, but not above that of multiple suggestions.
“Giving users the option to write instructions (i.e.non-diegetic prompts) increases the acceptance rate of suggestions for single suggestions, but not beyond that of multiple suggestions (top left).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataPeople mix diegetic and non-diegetic prompting, since single suggestions gained from both.
“Second, people combine diegetic and non-diegetic prompting, as single suggestions benefitted from both.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataTexts were somewhat shorter when written with single suggestions or with a UI that allowed instructions.
“Texts are slightly shorter when writing with single suggestions or with a UI allowing for instructions...”
- Also stated in: § 10 CONCLUSION · plain
“Writing time did not vary much and texts were slightly shorter with single suggestions and instructions (top center/right).”
Quote in the paperSubstantial claim- Also stated in: § 10 CONCLUSION · plain
Classified as probably not a substantial claim.
PremisecitationLonger phrase suggestions are perceived more as new ideas for writing.
“while longer phrase suggestions [6,26,37] are perceived more as new ideas for writing [1].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [1] source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationSuggestions can be triggered automatically after short inactivity, gated by a utility function, or explicitly requested with a hotkey.
“Related work showed suggestions automatically after short inactivity [4,6] or gated by a utility function [24].Alternatively, recent work has also explored designs in which users explicitly request suggestions with a hotkey [7,17,26,41,48].”
Quote in the paperNot a substantial claim0 of 8 sources checked- cites [4]
Not checked in the free run.
- cites [6]
Not checked in the free run.
- cites [24] source
Not checked in the free run.
- cites [7] source
Not checked in the free run.
- cites [17] source
Not checked in the free run.
- cites [26]
Not checked in the free run.
- cites [41] source
Not checked in the free run.
- cites [48]
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.
Study design and conductown dataThe analysis of open comments used a multi-step coding process with independent coding by two researchers and consensus resolution.
“We analysed the open comments from the final questionnaire in an approach adopting coding steps from Grounded Theory [10,29], in order to identify and report on the emerging aspects: First, two researchers inductively proposed codes for the data of 20 people.They then compared and clustered these codes to develop a common codebook.Then, they coded the first 20 plus 32 more participants and checked”
Quote in the paperNot a substantial claim0 of 2 sources checkedProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationProviding a few examples of inputs and outputs in a text prompt can help a model perform a task.
“Providing a few examples of inputs and outputs in such a text prompt can help the model solve a task [5,50].”
- Also stated in: § 2.1 Prompting in Large Language Models · plain
“Zhao et al. [50] show that providing a few examples of inputs and outputs can help to steer the model.”
Quote in the paperNot a substantial claim0 of 2 sources checked- cites [5] source
Not checked in the free run.
- cites [50]
Not checked in the free run.
- Also stated in: § 2.1 Prompting in Large Language Models · plain
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationSingle text suggestions can be shown inline, while multiple suggestions are shown as popup lists of about three to six entries.
“Single text suggestions can be shown inline [4,8,17,48], whereas multiple suggestions are shown as popup lists of about three to six entries [6,26].”
Quote in the paperNot a substantial claim0 of 6 sources checkedProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationOptimized prompts from prompt engineering are usually not meant for human consumption but are designed for LLMs to perform tasks effectively.
“However, optimized prompts constructed in the process of prompt engineering are usually not meant to be consumed by humans; rather, they are designed for LLMs to most effectively perform a task [28].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [28]
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable), S4 (based).
PremisereasoningFrom an HCI perspective, distinguishing diegetic from non-diegetic prompts helps identify patterns in how users perceive and interact with LLMs while writing.
“However, from an HCI perspective, this distinction allows us to identify patterns in the perception and interaction of users writing with LLMs.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Study design and conductnone givenThree researchers independently reviewed the stories and commented on their connection to the prompts, and one researcher reviewed all comments for consistency.
“To evaluate the degree to which participants engaged with the selected writing prompts during the user study (cf.Section 6), three researchers independently reviewed the stories and provided comments on their connection to the prompts.Finally, one researcher reviewed all comments to ensure consistency.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductown dataThe statistical models account for participants' individual differences and the type of chosen topics via random intercepts.
“The models account for participants' individual differences, as well as for the type of their chosen topics (creative story writing, argumentative writing), via random intercepts.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
PremisecitationDeep learning-based LLMs can solve several natural language processing tasks without task-specific training by using text prompts written in natural language.
“One primary advantage of Deep Learningbased LLMs is that they can solve several natural language processing tasks without being specifically trained on those.This can be done via text prompts written in natural language [5].”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [5] source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisereasoningTechnically, diegetic and non-diegetic prompts may be equivalent for an LLM because both are received as text input strings.
“Technically, there may not be a difference between diegetic and non-diegetic prompts for the LLM -both types are received by the model as text input strings.”
- Also stated in: § 2.3 Diegetic and Non-Diegetic Prompting in Existing Writing Interfaces · hedged
“From a technical perspective, for recent systems that use LLMs to generate text suggestions, there might be no difference between the user's text draft (i.e.diegetic text) and other text inputs to the language model”
Quote in the paperNot a substantial claim- Also stated in: § 2.3 Diegetic and Non-Diegetic Prompting in Existing Writing Interfaces · hedged
Probably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductnone givenThe order of the five UI variants was fully counterbalanced.
“The order of these five UIs was fully counterbalanced.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationOptimizing prompts is not trivial and requires extensive experience.
“However, optimizing prompts is not trivial and requires extensive experience [28].”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [28]
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.
Interpretationnone givenDespite potential misunderstanding of the 'shapeshifter' category, the majority of participants demonstrated attentiveness to the task and provided thoughtful reflections on the topic.
“Despite this potential misunderstanding, the majority of participants demonstrated attentiveness to the task and provided thoughtful reflections on the topic.”
Quote in the paperNot a substantial claimMay combine two claims.
InterpretationcitationDiegetic prompts do not require such shifts, although they still require engagement with displayed suggestions.
“In contrast, diegetic prompts do not require such shifts, although they still require engagement with displayed suggestions [4,6].”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [4]
Not checked in the free run.
- cites [6]
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Study design and conductreasoningChanges to the model over time are outside the authors' control, which restricts exact replication of studies of this kind.
“Potential changes to the model over time are beyond our control.This limits exact replicability for studies like this.”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown dataInstructions were situationally useful: participants commented on their benefits, used them in every fifth suggestion request, and experimented with different styles.
“At the same time, instructions were indeed (situationally) useful: Participants commented on their benefits (Section 8.2), used them in every fifth suggestion request, and experimented with different styles (Section 7.6).”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationA comparable pop-up is used in recent related work, though for suggestions in the middle of sentences rather than for instructions.
“A similar pop-up is used in recent related work (not for instructions but for suggestions in the middle of sentences; cf.[4]).”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [4]
Not checked in the free run.
May combine two claims.
Interpretationown dataMultiple suggestions and instructions were mentioned as contributing to feeling in control, matching the Likert results on control and influence.
“Overall, multiple suggestions and instructions were mentioned here as contributing to feeling in control, matching the Likert results on control and influence (𝑄 9 and 𝑄 10 in Figure 3).”
Quote in the paperNot a substantial claim- Rests on an assumption: The Likert questions validly measure the constructs they are intended to measure, such as control, distraction, and problems with instructions.
Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
Interpretationnone givenThere is a trend toward limited direct access to state-of-the-art LLMs for parts of the academic community.
“We see a trend of limited direct access to state-of-the-art LLMs for parts of the academic community, which is not easy to resolve.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationRecent text-to-image models let users produce images from text descriptions.
“Recent text to image models allow users to generate images from text descriptions [30,32,36].”
Quote in the paperNot a substantial claim0 of 3 sources checkedProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationChung et al. proposed a story ideation tool that guides an LLM through visual sketching.
“Chung et al. [9] proposed a new story ideation tool that uses visual sketching to guide a LLM.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [9]
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe amount of time participants spent writing in this study is comparable to that of related work.
“This is comparable to related work (e.g.11 min [26], 4 min [6], 10-12 min [48]).”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [26]
Not checked in the free run.
- cites [6]
Not checked in the free run.
- cites [48]
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductreasoningApart from the pre-study, the authors could not watch participants directly or ask them questions at interesting points in the interaction.
“However, we could not observe people directly or ask questions at interesting moments in the interaction, except for in our pre-study, which we used to refine our design.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationRecent studies, including this one, have used writing tasks drawn from writer communities and custom tasks.
“Pragmatically, writing tasks from writer communities and custom tasks have been used in recent studies (e.g.[6,26,41,48]), including ours.”
Quote in the paperNot a substantial claim0 of 4 sources checked- cites [6]
Not checked in the free run.
- cites [26]
Not checked in the free run.
- cites [41] source
Not checked in the free run.
- cites [48]
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Study design and conductnone givenThe authors had no direct access to the model and do not claim to have found the best settings for their use of it.
“We did not have access to the model directly and we do not claim to have identified the”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
InterpretationcitationIn Chung et al.'s tool the sketch becomes a text prompt, so the interaction is non-diegetic.
“Here the sketch is translated to a text prompt.This interaction is nondiegetic.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductreasoningThe instruction styles may be biased by the examples that were provided to participants.
“It is possible that the instruction styles (Section 7.6) are biased by the provided examples (Figure 8 in Appendix A).”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown dataPotential influences of the study setup do not necessarily work against instructions, or are less dominant than the effects of the UI design.
“This suggests that potential influences of the study setup do not necessarily work against instructions, or are less dominant than the effects of the UI design (e.g.sidebar vs integration at text cursor).”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Interpretationnone givenNone of the instruction styles go beyond what would also be a meaningful comment to a human co-author.
“At the moment, none of these styles go beyond what would also be a meaningful comment to a human co-author.”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown dataThe control offered by instructions was useful to guide single suggestions but was not better than having a choice of three suggestions to begin with.
“This indicates that the control offered by instructions was useful to guide single suggestions but not better than having a choice of three suggestions to begin with.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
InterpretationreasoningBecause the abstract shape is usually not part of the outcome, this interaction is treated as non-diegetic.
“Since the abstract shape is usually not part of the outcome we consider this interaction non-diegetic.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe image model converts the drawn shapes into a high-fidelity rendering.
“The image model then translates these shapes into a high-fidelity rendering.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [3] source
Not checked in the free run.
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationBau et al. give an example of guiding image models by painting shapes, such as a triangle standing for a mountain.
“Bau et al. [3] show an example of”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [3] source
Not checked in the free run.
Probably not a substantial claim: may not pass S2 (checkable), S4 (based).
Interpretationnone givenNo model, however good, can remove the cognitive and interaction costs of switching between diegetic and non-diegetic writing.
“even the best model cannot eliminate cognitive costs and interaction costs of switching between diegetic and non-diegetic writing.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
Interpretationnone givenDistinguishing the diegetic and non-diegetic perspectives lets researchers examine users' intentions and behavior when they interact with or design generative AI systems.
“Differentiating these two perspectives therefore allows researchers to analyse users' intention and behavior when interacting or designing systems with generative AI.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing), S4 (based).
Premisenone givenPrior systems in related work concentrated on non-diegetic prompts.
“While related work has presented systems that focused on non-diegetic prompts”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
InterpretationreasoningText-to-image prompts are non-diegetic because they are not part of the resulting visuals.
“These are non-diegetic prompts, because they are not part of the visuals.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationHa and Eck let users begin painting part of an image, providing diegetic information, and have the system continue or complete the painting.
“Ha and Eck [19] enable users to start painting a part of an image (i.e.providing diegetic information) and let the system continue or finish the painting.”
Quote in the paperNot a substantial claim0 of 1 sources checked- cites [19] source
Not checked in the free run.
The paper relies on this without support.
The 129 Prolific crowd workers are representative of the broader population of users who write with LLMs.
“We conducted a remote study with 129 crowd workers on Prolific, each writing five stories.”
The paper relies on this without support.
Suggestion acceptance rate is a valid indicator of users' preference for a suggestion UI.
“This is evident from highest acceptance rates with the multiple suggestions UI (Section 7.1)”
The paper relies on this without support.
The Likert questions validly measure the constructs they are intended to measure, such as control, distraction, and problems with instructions.
“matching the Likert results on control and influence (𝑄 9 and 𝑄 10 in Figure 3)”
The paper relies on this without support.
The qualitative coding categories and counts are reliable and meaningful.
“coded 36 times”
The paper relies on this without support.
The writing tasks used in the study are representative of short creative and argumentative writing tasks in general.
“Participants overall preferred choosing from multiple text suggestions presented to them, over actively writing instructions, in short creative and argumentative writing tasks.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
End-user studies suggest that prompting is challenging, with non-expert users making various errors and adopting ineffective strategies-a reflection of the demand on users' metacognitive monitoring and control.
“End-user studies suggest that prompting is challenging, with nonexpert users making various errors and adopting ineffective strategies-a reflection of the demand on users' metacognitive monitoring and control [33,41,43,85,103,174,205].”
From The Metacognitive Demands and Opportunities of Generative AI (2024) · cites this paper as [41]
Section 8.4 (page unavailable) reports that participants struggled to formulate instructions, and Section 8.3 describes one participant who had little success with instructions and changed approach. These passages support prompting being challenging for some non-expert users, but they do not document various user errors or establish ineffective strategies as a general finding. The paper also does not characterize these difficulties as demands on metacognitive monitoring and control. (LLM, conf=0.84)
Experimental — accuracy not yet measuredA more comprehensive understanding of views surrounding authenticity in human-AI co-creation remains elusive.
“Recent work on AI use for writing has begun to explore relevant constructs, such as ownership, authorship, and agency [11,13,39,49], but a more comprehensive understanding of views surrounding authenticity in human-AI co-creation remains elusive.”
From 'It was 80% me, 20% AI': Seeking Authenticity in Co-Writing with Large Language Models (2025) · cites this paper as [11]
Diegetic prompting is easier but affords less control than non-diegetic prompting, with preferences differing across users.
“For example, whereas diegetic prompting is easier, it affords less control than the alternative, with preferences differing across users [41].”
From The Metacognitive Demands and Opportunities of Generative AI (2024) · cites this paper as [41]
Dang et al. found that users preferred choosing from multiple suggestions and using content from the draft to guide LLM usage rather than writing external prompts which require more effort.
“Dang et al. [16] found that users preferred choosing from multiple suggestions and using content from the draft to guide LLM usage rather than writing external prompts which require more effort.”
From ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models (2024) · cites this paper as [16]
Recent HCI studies on LLM-based tool design have mainly focused on prompt engineering and exploring the generative capabilities of LLMs.
“Recent HCI studies on LLM-based tool design have mainly focused on prompt engineering [6,16,46,55,94,98] and exploring the generative capabilities of LLMs [50,58,63,67,97].”
From ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models (2024) · cites this paper as [16]
Prior studies show that while users prefer being able to choose from multiple options, having all of these options can negatively impact writing efficiency.
“This is in alignment with prior studies that show that while users prefer being able to choose from multiple options [16], having all of these options can negatively impact writing efficiency [9], indicating a need for option management support.”
From ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models (2024) · cites this paper as [16]
Dang et al. coined and distinguished between diegetic prompts that are part of draft with non-diegetic prompts that are external instructions to the LLM.
“Dang et al. [16] coined and distinguished between diegetic prompts that are part of draft with non-diegetic prompts that are external instructions to the LLM.”
From ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models (2024) · cites this paper as [16]
Users prefer to use the draft, i.e., diegetic information, to guide LLM prompts.
“and found that users prefer to use the draft, i.e., diegetic informtion, to guide LLM prompts.”
From ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models (2024) · cites this paper as [16]
Users may hesitate to write prompts when given multiple AI suggestions.
“Previous studies have shown that users may hesitate to write prompts when given multiple AI suggestions [16]”
From ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models (2024) · cites this paper as [16]
Interface elements can noticeably influence prompt-writing style.
“Our findings illustrate how interface elements can noticeably influence prompt-writing style, akin to previous work on, such as Dang et al. 's investigation of diegetic and non-diegetic prompts [16].”
From ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models (2024) · cites this paper as [16]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Suggestion-quality categories total more than the stated participants
MinorThe reported counts of negative, mixed, and positive assessments add to 19, although the paper says 18 participants commented on suggestion quality. The text does not say that any participant was counted in more than one category.
Reported 18 · recomputed 19
“Eighteen participants explicitly commented on suggestion quality: Eight were negative (𝑃 91 : "[...] had to edit most of it.").Nine felt suggestions were hit or miss (𝑃 130 : "Sometimes [the suggestions] helped, sometimes it didn't.").Two left positive comments”
Cites
R: A Language and Environment for Statistical Computing
R Core Team (2000) · doi:10.32614/r.manuals · Cited by 355,706 · see citing works
Fitting Linear Mixed-Effects Models Using <b>lme4</b>
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015) · Journal of Statistical Software · doi:10.18637/jss.v067.i01 · Cited by 86,835 · see citing works
<b>lmerTest</b> Package: Tests in Linear Mixed Effects Models
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017) · Journal of Statistical Software · doi:10.18637/jss.v082.i13 · Cited by 24,346 · see citing works
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372 · see citing works
Basics of Qualitative Research: Grounded Theory Procedures and Techniques
Long, D. R., Strauss, A., & Corbin, J. M. (1993) · Modern Language Journal · doi:10.2307/328955 · Cited by 17,612 · see citing works
Kaugeranna, R., Kaugeranna, E., & Aion, (. S. 4. 6. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.giscience.2023.43 · Cited by 13,695 · see citing works
Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., & Neubig, G. (2023) · ACM Computing Surveys · doi:10.1145/3560815 · Cited by 4,361 · see citing works
Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., & Neubig, G. (2021) · arXiv · doi:10.48550/arxiv.2107.13586 · Cited by 4,352 · cited 2× · see citing works
“However, optimizing prompts is not trivial and requires extensive experience [28].”
BackgroundHierarchical Text-Conditional Image Generation with CLIP Latents
Ramesh, A., Dhariwal, P., Nichol, A., Chu, C., & Chen, M. (2022) · arXiv · doi:10.48550/arxiv.2204.06125 · Cited by 2,511 · see citing works
Language Models as Knowledge Bases?
Petroni, F., Rocktäschel, T., Riedel, S., Lewis, P., Bakhtin, A., Wu, Y., & Miller, A. (2019) · Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) · doi:10.18653/v1/d19-1250 · Cited by 1,834 · see citing works
Enhancing the explanatory power of usability heuristics
Nielsen, J. (1994) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/191666.191729 · Cited by 1,439 · see citing works
GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models
Nichol, A., Dhariwal, P., Ramesh, A., Shyam, P., Mishkin, P., McGrew, B., Sutskever, I., & Chen, M. (2021) · arXiv · doi:10.48550/arxiv.2112.10741 · Cited by 1,357 · see citing works
Exploiting Cloze-Questions for Few-Shot Text Classification and Natural Language Inference
Schick, T., & Schütze, H. (2021) · Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume · doi:10.18653/v1/2021.eacl-main.20 · Cited by 1,305 · see citing works
StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery
Patashnik, O., Wu, Z., Shechtman, E., Cohen‐Or, D., & Lischinski, D. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00209 · Cited by 1,079 · see citing works
How Can We Know What Language Models Know?
Jiang, Z., Xu, F. F., Araki, J., & Neubig, G. (2020) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00324 · Cited by 1,002 · see citing works
Modeling and remodeling writing
Hayes, J. R. (2012) · Written Communication · doi:10.1177/0741088312451260 · Cited by 893 · see citing works
A Neural Representation of Sketch DrawingsPreprint
Ha, D., & Eck, D. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1704.03477 · Cited by 538 · see citing works
Wu, T., Terry, M. A., & Cai, C. J. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3517582 · Cited by 536 · see citing works
Calibrate Before Use: Improving Few-Shot Performance of Language Models
Zhao, T. Z., Wallace, E., Feng, S., Klein, D., & Singh, S. (2021) · arXiv · doi:10.48550/arxiv.2102.09690 · Cited by 513 · see citing works
GAN Dissection: Visualizing and Understanding Generative Adversarial NetworksPreprint
Bau, D., Zhu, J.-Y., Strobelt, H., Zhou, B., Tenenbaum, J. B., Freeman, W. T., & Torralba, A. B. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.10597 · Cited by 404 · see citing works
Wordcraft: Story Writing With Large Language Models
Yuan, A., Coenen, A., Reif, E., & Ippolito, D. (2022) · 27th International Conference on Intelligent User Interfaces · doi:10.1145/3490099.3511105 · Cited by 374 · see citing works
BARTScore: Evaluating Generated Text as Text Generation
Yuan, W., Neubig, G., & Liu, P. (2021) · arXiv · doi:10.48550/arxiv.2106.11520 · Cited by 353 · see citing works
Lee, M., Liang, P., & Yang, Q. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3502030 · Cited by 345 · cited 8× · see citing works
“Single text suggestions can be shown inline [4, 8, 17, 48], whereas multiple suggestions are shown as popup lists of about three to six entries [6, 26].”
BackgroundResultInfluentialTaleBrush: Sketching Stories with Generative Pretrained Language Models
Chung, J. J. Y., Kim, W., Yoo, K. M., Lee, H., Adar, E., & Chang, M. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3501819 · Cited by 206 · see citing works
Interactive and Visual Prompt Engineering for Ad-hoc Task Adaptation With Large Language Models
Strobelt, H., Webson, A., Sanh, V., Hoover, B., Beyer, J., Pfister, H., & Rush, A. M. (2022) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2022.3209479 · Cited by 204 · see citing works
Interactive and Visual Prompt Engineering for Ad-hoc Task Adaptation with Large Language Models
Strobelt, H., Webson, A., Sanh, V., Hoover, B., Beyer, J., Pfister, H., & Rush, A. M. (2022) · arXiv · doi:10.48550/arxiv.2208.07852 · Cited by 204 · see citing works
Goel, M., Findlater, L., & Wobbrock, J. O. (2012) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/2207676.2208662 · Cited by 203 · see citing works
PromptChainer: Chaining Large Language Model Prompts through Visual Programming
Wu, T., Jiang, E., Donsbach, A., Gray, J., Molina, A., Terry, M., & Cai, C. J. (2022) · CHI Conference on Human Factors in Computing Systems Extended Abstracts · doi:10.1145/3491101.3519729 · Cited by 198 · see citing works
Kannan, A., Kurach, K., Ravi, S., Kaufmann, T., Tomkins, A., Miklos, B., Corrado, G. S., Lukacs, L., Ganea, M., Young, P., et al. (2016) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/2939672.2939801 · Cited by 182 · see citing works
Gmail Smart Compose: Real-Time Assisted Writing
Chen, M. X., Lee, B. N., Bansal, G., Cao, Y., Zhang, S., Lu, J., Tsay, J., Wang, Y., Dai, A. M., Chen, Z., et al. (2019) · Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3292500.3330723 · Cited by 169 · see citing works
Sparks: Inspiration for Science Writing using Language Models
Gero, K. I., Liu, V., & Chilton, L. (2022) · Designing Interactive Systems Conference · doi:10.1145/3532106.3533533 · Cited by 161 · see citing works
PromptMaker: Prompt-based Prototyping with Large Language Models
Jiang, E., Olson, K., Toh, E., Molina, A., Donsbach, A., Terry, M., & Cai, C. J. (2022) · CHI Conference on Human Factors in Computing Systems Extended Abstracts · doi:10.1145/3491101.3503564 · Cited by 157 · cited 1× · see citing works
“Similarily, PromptMaker [22] allows users to prototype new AI functionalities using language prompts.”
BackgroundWatchWriter: Tap and Gesture Typing on a Smartwatch Miniature Keyboard with Statistical Decoding
Gordon, M., Ouyang, T., & Zhai, S. (2016) · Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems · doi:10.1145/2858036.2858242 · Cited by 150 · see citing works
What Makes Good In-Context Examples for GPT-$3$?
Liu, J., Shen, D., Zhang, Y., Dolan, B., Carin, L., & Chen, W. (2021) · arXiv · doi:10.48550/arxiv.2101.06804 · Cited by 148 · see citing works
Effects of Language Modeling and its Personalization on Touchscreen Typing Performance
Fowler, A., Partridge, K., Chelba, C., Bi, X., Ouyang, T., & Zhai, S. (2015) · Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems · doi:10.1145/2702123.2702503 · Cited by 127 · see citing works
Where to Hide a Stolen Elephant: Leaps in Creative Writing with Multimodal Machine Intelligence
Singh, N., Bernal, G., Savchenko, D., & Glassman, E. L. (2023) · ACM Transactions on Computer-Human Interaction · doi:10.1145/3511599 · Cited by 119 · see citing works
Buschek, D., Zürn, M., & Eiband, M. (2021) · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445372 · Cited by 108 · see citing works
Self-citationA Cost-Benefit Study of Text Entry Suggestion Interaction
Quinn, P., & Zhai, S. (2016) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/2858036.2858305 · Cited by 102 · see citing works
Dunlop, M., & Levine, J. (2012) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/2207676.2208659 · Cited by 99 · see citing works
Goel, M., Jansen, A., Mandel, T., Patel, S., & Wobbrock, J. O. (2013) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/2470654.2481386 · Cited by 96 · see citing works
RPackagemultgee: A Generalized Estimating Equations Solver for Multinomial Responses
Touloumis, A. (2015) · Journal of Statistical Software · doi:10.18637/jss.v064.i08 · Cited by 90 · see citing works
BERTese: Learning to Speak to BERT
Haviv, A., Berant, J., & Globerson, A. (2021) · Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume · doi:10.18653/v1/2021.eacl-main.316 · Cited by 84 · see citing works
Creative Help: A Story Writing Assistant
Roemmele, M., & Gordon, A. S. (2015) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-27036-4_8 · Cited by 83 · see citing works
Complementing Text Entry Evaluations with a Composition Task
Vertanen, K., & Kristensson, P. O. (2014) · ACM Transactions on Computer-Human Interaction · doi:10.1145/2555691 · Cited by 62 · see citing works
Story Centaur: Large Language Model Few Shot Learning as a Creative Writing Tool
Swanson, B., Mathewson, K. W., Pietrzak, B., Chen, S., & Dinalescu, M. (2021) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2021.eacl-demos.29 · Cited by 56 · see citing works
PEER: A Collaborative Language Model
Schick, T., Dwivedi-Yu, J., Jiang, Z., Petroni, F., Lewis, P., Izacard, G., You, Q., Nalmpantis, C., Grave, E., & Riedel, S. (2022) · arXiv · doi:10.48550/arxiv.2208.11663 · Cited by 49 · see citing works
Bhat, A., Agashe, S., Oberoi, P., Mohile, N., Jangir, R., & Joshi, A. (2023) · Proceedings of the 28th International Conference on Intelligent User Interfaces · doi:10.1145/3581641.3584060 · Cited by 47 · see citing works
Bhat, A., Agashe, S., Mohile, N., Oberoi, P., Jangir, R., & Joshi, A. (2022) · arXiv · doi:10.48550/arxiv.2208.00636 · Cited by 47
How Novelists Use Generative Language Models: An Exploratory User Study.
Calderwood, A., Qiu, V., Gero, K. I., & Chilton, L. B. (2020) · Cited by 35 · see citing works
Sentiment Bias in Predictive Text Recommendations Results in Biased Writing
Arnold, K., Chauncey, K., & Gajos, K. (2018) · Canadian Human-Computer Communications Society / Société canadienne du dialogue humain-machine · doi:10.20380/gi2018.07 · Cited by 31 · see citing works
A Design Space for Writing Support Tools Using a Cognitive Process Model of Writing
Gero, K., Calderwood, A., Li, C., & Chilton, L. (2022) · Proceedings of the First Workshop on Intelligent and Interactive Writing Assistants (In2Writing 2022) · doi:10.18653/v1/2022.in2writing-1.2 · Cited by 25 · see citing works
Grounded Theory Method in Human-Computer Interaction and Computer-Supported Cooperative Work
Muller, M. J., & Kogan, S. (2012) · CRC Press eBooks · doi:10.1201/b11963-ch-44 · Cited by 10 · see citing works
Flagged first, then by citations
Research works printed but not linked: 3
- Steven Goodman, Erin Buehler, Patrick Clary, Andy Coenen, Aaron Michael Donsbach, Tiffanie Horne, Michal Lahav, Bob MacDonald, Rain Breaw Michaels, Ajit Narayanan, Mahima Pushkarna, Joel Christopher Riley, Alex Santana, Lei Shi, Rachel Sweeney, Phil Weaver, Ann Yuan, and Meredith Ringel Morris. 2022. LaMPost: Evaluation of an AI-assisted Writing Email Editor Prototype for Adults with Dyslexia. https://arxiv.org/abs/2207.02308 · arXiv
- Herbert Alexander Simon. 1996. The sciences of the artificial(3. ed. ed.). MIT Press, Cambridge, Mass.
- 10.18653/v1 · DOI
Cited by
Sorted by citations
The Metacognitive Demands and Opportunities of Generative AI
Tankelevitch, L., Kewenig, V., Simkute, A., Scott, A. E., Sarkar, A., Sellen, A., & Rintel, S. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642902 · Cited by 280 · see citing works
Prompting AI Art: An Investigation into the Creative Skill of Prompt Engineering
Oppenlaender, J., Linder, R., & Silvennoinen, J. M. (2024) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2024.2431761 · Cited by 308 · see citing works
Harnessing the power of AI in qualitative research: Exploring, using and redesigning ChatGPT
Zhang, H., Wu, C., Xie, J., Lyu, Y., Cai, J., & Carroll, J. M. (2025) · Computers in Human Behavior Artificial Humans · doi:10.1016/j.chbah.2025.100144 · Cited by 290 · see citing works
A Design Space for Intelligent and Interactive Writing Assistants
Lee, M., Gero, K. I., Chung, J. J. Y., Shum, S. B., Raheja, V., Shen, H., Venugopalan, S., Wambsganss, T., Zhou, D., Alghamdi, E. A., et al. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642697 · Cited by 136 · see citing works
“Finally, most ecosystem dimensions are, as previously noted, under-represented, representing a rich area of future study as writing assistants become more widely adopted and the circumstance and context of their adoption becomes increasingly important.”
InfluentialSelf-citationHow Knowledge Workers Think Generative AI Will (Not) Transform Their Industries
Woodruff, A., Shelby, R. M., Kelley, P. G., Rousso-Schindler, S., Smith-Loud, J., & Wilcox, L. G. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642700 · Cited by 126 · see citing works
DirectGPT: A Direct Manipulation Interface to Interact with Large Language Models
Masson, D., Malacria, S., Casiez, G., & Vogel, D. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642462 · Cited by 117 · see citing works
Qin, H. X., Jin, S., Gao, Z., Fan, M., & Hui, P. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642105 · Cited by 82 · see citing works
Ironies of Generative AI: Understanding and Mitigating Productivity Loss in Human-AI Interaction
Simkute, A., Tankelevitch, L., Kewenig, V., Scott, A. E., Sellen, A., & Rintel, S. (2024) · International Journal of Human–Computer Interaction · doi:10.1080/10447318.2024.2405782 · Cited by 89 · see citing works
“This not only makes prompting a challenge (e.g., Dang et al., 2023; Zamfirescu-Pereira et al., 2023) but also the evaluation of outputs (e.g., Barke et al., 2023; Liang et al., 2023; Weisz et al., 2021) as the two are inextricably intertwined in current systems.”
InfluentialThe Value, Benefits, and Concerns of Generative AI-Powered Assistance in Writing
Li, Z., Liang, C., Peng, J., & Yin, M. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642625 · Cited by 81 · see citing works
Generative AI in the Wild: Prospects, Challenges, and Strategies
Sun, Y., Jang, E., Ma, F., & Wang, T. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642160 · Cited by 64 · see citing works
A Taxonomy for Human-LLM Interaction Modes: An Initial Exploration
Gao, J., Gebreegziabher, S. A., Choo, K. T. W., Li, T. J.-J., Perrault, S. T., & Malone, T. W. (2024) · Extended Abstracts of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613905.3650786 · Cited by 57 · see citing works
Pang, R. Y., Schroeder, H., Smith, K. S., Barocas, S., Xiao, Z., Tseng, E., & Bragg, D. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713726 · Cited by 68 · see citing works
Reza, M., Laundry, N. M., Musabirov, I., Dushniku, P., Yu, Z. Y. “., Mittal, K., Grossman, T., Liut, M., Kuzminykh, A., & Williams, J. J. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3641899 · Cited by 52 · see citing works
DATATALES: Investigating the use of Large Language Models for Authoring Data-Driven Articles
Sultanum, N., & Srinivasan, A. (2023) · 2023 IEEE Visualization and Visual Analytics (VIS) · doi:10.1109/vis54172.2023.00055 · Cited by 52 · see citing works
Tanprasert, T., Fels, S., Sinnamon, L. S., & Yoon, D. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642513 · Cited by 52 · see citing works
Bhattacharjee, A., Zeng, Y.-C., Xu, S., Kulzhabayeva, D., Ma, M., Kornfield, R., Ahmed, S. I., Mariakakis, A., Czerwinski, M. P., Kuzminykh, A., et al. (2023) · International Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642081 · Cited by 59 · see citing works
'It was 80% me, 20% AI': Seeking Authenticity in Co-Writing with Large Language Models
Hwang, A. H.-C., Liao, Q. V., Blodgett, S. L., Olteanu, A., & Trischler, A. (2025) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3711020 · Cited by 42 · see citing works
Drosos, I., Sarkar, A., Xu, X., Negreanu, C., Rintel, S., & Tankelevitch, L. (2024) · Proceedings of the 3rd Annual Meeting of the Symposium on Human-Computer Interaction for Work · doi:10.1145/3663384.3663389 · Cited by 30 · see citing works
Co-Writing with AI, on Human Terms: Aligning Research with User Demands Across the Writing Process
Reza, M., Thomas-Mitchell, J., Dushniku, P., Laundry, N., Williams, J. J., & Kuzminykh, A. (2025) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3757566 · Cited by 37 · see citing works
Zhong, S., Huang, Z., Gao, S., Wen, W., Lin, L., Žitnik, M., & Zhou, P. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.01258 · Cited by 30 · see citing works
A Map of Exploring Human Interaction Patterns with LLM: Insights into Collaboration and Creativity
Li, J., Li, J., & Su, Y. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-60615-1_5 · Cited by 23 · see citing works
Amuse: Human-AI Collaborative Songwriting with Multimodal Inspirations
Kim, Y., Lee, S.-J., & Donahue, C. (2025) · Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713818 · Cited by 23 · see citing works
Zhou, Y., & Chen, C. (2025) · Design and Artificial Intelligence · doi:10.1016/j.daai.2025.100017 · Cited by 23 · see citing works
Authors' Values and Attitudes Towards AI-bridged Scalable Personalization of Creative Language Arts
Kim, T., Han, H., Adar, E., Kay, M., & Chung, J. J. Y. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642529 · Cited by 20 · see citing works
Chen, Q., Shuai, W., Zhang, J., Sun, Z., & Cao, N. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642480 · Cited by 20 · see citing works
DBox: Scaffolding Algorithmic Programming Learning through Learner-LLM Co-Decomposition
Ma, S., Wang, J., Zhang, Y., Ma, X., & Wang, A. Y. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713748 · Cited by 20 · see citing works
Ortloff, A.-M., Martius, F., Meier, M., Sans-Raimbault, T., Geierhaas, L., & Smith, M. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713671 · Cited by 19 · see citing works
Siddiqui, M. N., Pea, R. D., & Subramonyam, H. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714119 · Cited by 18 · see citing works
Collage is the New Writing: Exploring the Fragmentation of Text and User Interfaces in AI Tools
Buschek, D. (2024) · Designing Interactive Systems Conference · doi:10.1145/3643834.3660681 · Cited by 20 · see citing works
Self-citationDing, Z., Smith, A., Zhang, W., Tetreault, J., & Jaimes, A. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.217 · Cited by 20 · see citing works
FusAIn: Composing Generative AI Visual Prompts Using Pen-based Interaction
Peng, X., Koch, J., & Mackay, W. E. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714027 · Cited by 15 · see citing works
Do LLMs Meet the Needs of Software Tutorial Writers? Opportunities and Design Implications
Bhat, A., Shrivastava, D., & Guo, J. (2024) · Designing Interactive Systems Conference · doi:10.1145/3643834.3660692 · Cited by 13 · see citing works
Making the Write Connections: Linking Writing Support Tools with Writer Needs
Zhao, Z., Masson, D., Kim, Y., Penn, G., & Chevalier, F. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713161 · Cited by 14 · see citing works
ChatTL;DR – You Really Ought to Check What the LLM Said on Your Behalf
Gould, S. J. J., Brumby, D. P., & Cox, A. L. (2024) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3613905.3644062 · Cited by 10 · see citing works
CoCo Matrix: Taxonomy of Cognitive Contributions in Co-writing with Intelligent Agents
Wan, R., Gebreegziabher, S. A., Li, T. J.-J., & Badillo-Urquiola, K. (2024) · Creativity and Cognition · doi:10.1145/3635636.3664260 · Cited by 9 · see citing works
SketchGPT: A Sketch-based Multimodal Interface for Application-Agnostic LLM Interaction
Huang, Z., Gao, C., Shan, Y., Hu, H., Li, Q., Deng, X., Ma, C., Lai, Y., Liu, Y., Tian, F., et al. (2025) · ACM Symposium on User Interface Software and Technology (UIST) · doi:10.1145/3746059.3747598 · Cited by 8 · see citing works
Maier, S., Schneider, M., & Feuerriegel, S. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791185 · Cited by 8 · see citing works
Overall Prompting Effectiveness for Optimising Human-Machine Interaction in Cyber-Physical Systems
Wilbers, S., Espinosa-Leal, L., van de Sand, R., & Reiff-Stephan, J. (2023) · Journal of Integrated Design and Process Science · doi:10.1177/10920617241295859 · Cited by 7 · see citing works
Visual Story-Writing: Writing by Manipulating Visual Representations of Stories
Masson, D., Zhao, Z., & Chevalier, F. (2025) · ACM Symposium on User Interface Software and Technology (UIST) · doi:10.1145/3746059.3747758 · Cited by 6 · see citing works
Amin, R. M., Kühle, O. H., Buschek, D., & Butz, A. M. (2025) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3706599.3720243 · Cited by 6 · see citing works
Self-citationSchömbs, S., Zhang, Y., Gonçalves, J., & Johal, W. (2025) · International Conference on Human-Agent Interaction (HAI) · doi:10.1145/3765766.3765795 · Cited by 6 · see citing works
FlowGPT: Exploring Domains, Output Modalities, and Goals of Community-Generated AI Chatbots
Li, X., Han, Y., Liu, D., An, P., & Niu, S. (2024) · ACM Conference on Computer Supported Cooperative Work (CSCW) · doi:10.1145/3678884.3681875 · Cited by 5 · see citing works
Authoring LLM-Based Assistance for Real-World Contexts and Tasks
Dang, H., Lafreniere, B., Grossman, T., Todi, K., & Li, M. (2025) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3708359.3712164 · Cited by 5 · see citing works
Self-citationInk and algorithm: Exploring temporal dynamics in generative <scp>AI</scp> ‐assisted writing
Yang, K., Cheng, Y., Zhao, L., Raković, M., Swiecki, Z., Gašević, D., & Chen, G. (2026) · British Journal of Educational Technology · doi:10.1111/bjet.70031 · Cited by 4 · see citing works
Investigating How Computer Science Researchers Design Their Co-Writing Experiences With AI
Roffarello, A. M., Calò, T., Scibetta, L., & Russis, L. D. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713205 · Cited by 5 · see citing works
Ma, Z., Mei, Y., Bruderlein, C., Gajos, K. Z., & Pan, W. (2025) · Annual Symposium on Human-Computer Interaction for Work (CHIWORK) · doi:10.1145/3729176.3729194 · Cited by 6 · see citing works
Kim, Y., Le, T.-L. V., Kim, D., Lee, M., & Lee, S.-J. (2025) · Designing Interactive Systems Conference · doi:10.1145/3715336.3735740 · Cited by 7 · see citing works
Exploring Mobile Touch Interaction with Large Language Models
Zindulka, T., Sekowski, J. M., Lehmann, F., & Buschek, D. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713554 · Cited by 4 · see citing works
Self-citationCorpusStudio: Surfacing Emergent Patterns In A Corpus Of Prior Work While Writing
Dang, H., Swoopes, C., Buschek, D., & Glassman, E. L. (2025) · International Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713974 · Cited by 13 · see citing works
Self-citationSandTouch: Empowering Virtual Sand Art in VR with AI Guidance and Emotional Relief
Liu, L., Ren, J., Fan, Z., Li, C., He, G., Wang, C., Gao, Y., & Li, C. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714275 · Cited by 5 · see citing works
Zindulka, T., Goller, S., Lehmann, F., & Buschek, D. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713890 · Cited by 3 · see citing works
Self-citationExamining Human-AI Collaboration for Co-Writing Constructive Comments Online
Shahid, F., Dittgen, M., Naaman, M., & Vashistha, A. (2025) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3757591 · Cited by 3 · see citing works
Exploring Direct Instruction and Summary-Mediated Prompting in LLM-Assisted Code Modification
Tang, N., Smith, E., Huang, Y., Collin, M., & Li, T. J.-J. (2025) · IEEE Symposium on Visual Languages and Human-Centric Computing (VL/HCC) · doi:10.1109/vl-hcc65237.2025.00017 · Cited by 2 · see citing works
Pang, R. Y., Feng, K. J. K., Feng, S., Li, C., Shi, W., Tsvetkov, Y., Heer, J., & Reinecke, K. (2026) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3742413.3789091 · Cited by 12 · see citing works
Epic-Level Text Generation with LLM through Auto-prompted Reinforcement Learning
Qi, Q., Ni, L., Wang, Z., Zhang, L., Liu, J., & Witbrock, M. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10650358 · Cited by 2 · see citing works
Chen, Y., Xiao, Y., Zinat, K. T., Yamashita, N., Gao, G., & Liu, Z. (2025) · Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713693 · Cited by 2 · see citing works
Fitzsimons, A., Gerber, E. M., & Long, D. (2024) · Workshop on Intelligent and Interactive Writing Assistants (In2Writing) · doi:10.1145/3690712.3690719 · Cited by 1 · see citing works
Augmenting Human Cognition Through Generative AI
Rogers, Y., Reicherts, L., Zhang, Z. T., & Hassib, M. (2026) · Springer eBooks · doi:10.1007/978-3-031-61050-9_10-1 · Cited by 1 · see citing works
Hybrid interpretation with generative AI: a pilot study using the A Duo with ChatGPT instrument
Esteban, M. P. S. (2026) · AI & Society · doi:10.1007/s00146-025-02839-0 · Cited by 2 · see citing works
“Control Is a Trajectory, Not a Point”: Conceptualizing Control in Human-AI Co-Creativity
Issak, A., Rezwana, J., & Harteveld, C. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3790861 · Cited by 1 · see citing works
Generative Artificial Intelligence for the Visualization of Source Code as Comics
Heidrich, D., Schreiber, A., & Theis, S. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-60114-9_4 · Cited by 1 · see citing works
How different mental models of AI-based writing assistants impact writers’ interactions with them
Rismani, S., Blodgett, S. L., Olteanu, A., Liao, Q. V., & Moon, A. (2024) · Workshop on Intelligent and Interactive Writing Assistants (In2Writing) · doi:10.1145/3690712.3690722 · Cited by 1 · see citing works
Interacting with AI by Manipulating Intents
Kim, T. S. (2025) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3706599.3707611
Lehmann, F., & Buschek, D. (2025) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3733053
Self-citationvan As, J. (2025) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3373-1414-3.ch003
MetaPipe: Predicting Metaphoric Associations and Turning Your Metaphoric Imagination into Reality
Qian, W., Hu, Z., Hu, J., Li, J., & Hong, R. (2025) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2025.2581257
Jahna, R. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.5842662
Developer Prompts in Practice: An Empirical Study of Bias, Security, and Optimization
Rzig, D. E., Paul, D. J., Pister, K. S., Henkel, J. M., & Hassan, F. (2025) · Proceedings of the ... ACM-IEEE International Symposium on Empirical Software Engineering and Measurement/International Symposium on Empirical Software Engineering and Measurement · doi:10.1109/esem64174.2025.00067
LAPPI: Interactive Optimization With LLM-Assisted Preference-Based Problem Instantiation
Kuroki, S., Nakagawa, M., Yoshida, S., Koyama, Y., & Kozuno, T. (2026) · IEEE Access · doi:10.1109/access.2026.3679052
Wang, Z., Zhang, G., Chen, X., Hao, S., Li, Y., Xiong, G., Yang, Z., Yang, K., Sim, M. Y., & Xu, K. (2026) · Synthesis lectures on human language technologies · doi:10.1007/978-3-032-06264-2_11
Yin, Y., Wu, M., Huang, R., Tong, X., Zhou, J., Yu, C., & Shi, Y. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3790848
From Use to Oversight: How Mental Models Influence User Behavior and Output in AI Writing Assistants
Rismani, S., Blodgett, S. L., Liao, Q., Olteanu, A., & Moon, A. (2026) · International Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791670 · Cited by 2 · see citing works
Sci-Fi Spark: A Human-AI Co-Creation System for Science Fiction Ideation
Jiang, Z., Cheang, W., Xu, X., Zuo, H., Zheng, C., & Chen, L. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791950
Everyone is a product manager: MAUX, a multi-agent framework for democratized user experience design
Sun, Y., Li, F., Chang, D., & Zhang, Z. (2026) · Advanced Engineering Informatics · doi:10.1016/j.aei.2026.104715
Supporting Reviewing Reviews: How HCI Authors Handle Peer Reviews of Manuscripts
Yeung, C. A., Stark, J., Li, J., Chevalier, F., Park, J., Kim, Y., & Tang, A. (2025) · International Conference on Human-Engaged Computing (ICHEC) · doi:10.1145/3786995.3786998
Petridis, S., Liu, M. X., Fiannaca, A. J., Cai, C. J., & Terry, M. A. (2026) · Designing Interactive Systems Conference · doi:10.1145/3800645.3812956
Hwang, S., Hwang, S., Lee, S., Kim, D., & Lee, K. (2026) · Symposium on Designing Interactive Systems · doi:10.1145/3800645.3812859 · Cited by 4 · see citing works
Zhang, X., & Lee, J. (2026) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2026.2690096
The meek shall inherit the Earth: personality, empathy, and the human shaping of LLMs
Mangalam, M. (2026) · AI and Ethics · doi:10.1007/s43681-026-01215-y
Speaking with LLMs: Co-Designing Voice Interaction Patterns with Blind and Visually Impaired Users
Pucci, E., Fede, G. D., Gilbert, M., Leporini, B., Andolina, S., & Matera, M. (2026) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2026.2688501
Critical Thinking Patterns and Their Associations with Creative Performance in Human-AI Co-Creation
Chen, J., & Shi, Y. (2026) · Journal of Data and Information Science · doi:10.1515/jdis-2025-0487
Zindulka, T., Sekowski, J. M., Benharrak, K., & Buschek, D. (2026) · Human-Computer Interaction · doi:10.1080/07370024.2026.2713509
Self-citationEchoWeave: Designing an AI-Driven Conversational Agent System for Product Design Research Interviews
Chen, X., Zhang, L., Xu, L., Bao, L., García, C., & Sareh, P. (2026) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2026.2744496
Performance and Metacognition Disconnect when Reasoning in Human-AI Interaction
Fernandes, D., Villa, S., Nicholls, S., Haavisto, O., Buschek, D., Schmidt, A., Kosch, T., Shen, C., & Welsch, R. (2024) · arXiv · doi:10.48550/arxiv.2409.16708 · Cited by 71
Self-citation
- Authors
- Dang, H., Goller, S., Lehmann, F., & Buschek, D.
- Affiliations
- University of Bayreuth (DE)
- Venue
- Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems
- Publisher
- ACM
- Published
- 2023-04-19
- Pages
- 1-17
- OpenAlex
- W4323570570
- Semantic Scholar
- fccf8776d7525627c518a56a1f4db367a4d7120b
- Cited by
- 83
- References
- 48
- Citations by year
- 2026: 21 · 2025: 32 · 2024: 27 · 2023: 3
- FWCI
- 14.69
- Type
- conference paper
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- CHI Conference on Human Factors in Computing Systems (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Bavarian State Ministry of Science and the Arts
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Speech and dialogue systems
- Physical Sciences › Computer Science › Human-Computer Interaction › Innovative Human-Technology Interaction
- Physical Sciences › Computer Science › Human-Computer Interaction › Usability and User Interface Design
- Keywords
- large language models, prompting, user interface design, human-AI interaction, generative AI, text generation, user choice, user control, GPT-3, crowd workers
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-04.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-04.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
