Will AI Console Me when I Lose my Pet? Understanding Perceptions of AI-Mediated Email Writing
Open accessAbstract
Large language models are increasingly mediating, modifying, and even generating messages for users, but the receivers of these messages may not be aware of the involvement of AI. To examine this emerging direction of AI-Mediated Communication (AI-MC), we investigate people’s perceptions of AI written messages. We analyze how such perceptions change in accordance with the interpersonal emphasis of a given message. We conducted both large-scale surveys and in-depth interviews to investigate how a diverse set of factors influence people’s perceived trust in AI-mediated writing of emails. We found that people’s trust in email writers decreased when they were told that AI was involved in the writing process. Surprisingly trust increased when AI was used for writing more interpersonal emails (as opposed to more transactional ones). Our study provides insights regarding how people perceive AI-MC and has practical design implications on building AI-based products to aid human interlocutors in communication1.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
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?
The evaluation criteria (likeability, enjoyment, creativity) were chosen because they were used in prior work to assess AI-assisted writing.
“We chose these evaluation criteria as they were also used in prior work to assess AI-assisted writing (e.g., [32,45]).”
From 'It was 80% me, 20% AI': Seeking Authenticity in Co-Writing with Large Language Models (2025) · cites this paper as [45]
The Methods section studies perceptions of AI-mediated email writing, so the paper is relevant prior work. But the Dependent Variables section identifies ability, benevolence, and integrity as its evaluation dimensions; the cited passages do not show it using likeability, enjoyment, or creativity. (LLM, conf=0.82)
Experimental — accuracy not yet measuredPrevious AIMC work has found readers to hold negative opinions about messages and their writers once they become aware of the message being written with AI assistance.
“This is a surprising result given that previous AIMC work has found readers to hold negative opinions about messages and their writers once they become aware of the message being written with AI assistance [31,44,59].”
From 'It was 80% me, 20% AI': Seeking Authenticity in Co-Writing with Large Language Models (2025) · cites this paper as [44]
The Introduction reports lower trust in emails perceived as AI-mediated and negative judgments of senders when AI was proposed for highly interpersonal messages. The claim is broader than the study: its emails were human-written but presented as AI-mediated, and the strongest negative judgments of writers concerned highly interpersonal messages. (LLM, conf=0.86)
Experimental — accuracy not yet measuredLiu et al. [66] found that perceived trustworthiness of AI-only communication was rated lowest regardless of interpersonal emphasis.
“The researchers found that, regardless of interpersonal emphasis, that the perceived trustworthiness of AI-only communication was rated the lowest.”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
The citation lists haven’t been fetched yet
References 35 · Cited by 92
Research works printed but not linked: 17
- Ritu Agarwal and Jayesh Prasad. 1999. Are individual differences germane to the acceptance of new information technologies? Decision sciences 30, 2 (1999), 361-391.
- Joyce Berg, John Dickhaut, and Kevin McCabe. 1995. Trust, reciprocity, and social history. Games and economic behavior 10, 1 (1995), 122-142.
- Karen S Cook, Toshio Yamagishi, Coye Cheshire, Robin Cooper, Masafumi Mat- suda, and Rie Mashima. 2005. Trust building via risk taking: A cross-societal experiment. Social psychology quarterly 68, 2 (2005), 121-142.
- Jens Edlund, Joakim Gustafson, Mattias Heldner, and Anna Hjalmarsson. 2008. Towards human-like spoken dialogue systems. Speech communication 50, 8-9 (2008), 630-645.
- Amanda J Ferguson and Randall S Peterson. 2015. Sinking slowly: Diversity in propensity to trust predicts downward trust spirals in small groups. Journal of Applied Psychology 100, 4 (2015), 1012.
- Susan C Herring. 1996. Computer-mediated communication: Linguistic, social, and cross-cultural perspectives. Vol. 39. John Benjamins Publishing.
- Matthew Hines. 2019. I smell a bot: California's SB 1001, free speech, and the future of bot regulation. Hous. L. Rev. 57 (2019), 405.
- Jess Hohenstein and Malte Jung. 2018. AI-Supported Messaging: An Investigation of Human-Human Text Conversation with AI Support (CHI EA '18). Association for Computing Machinery, New York, NY, USA, 1-6. https://doi.org/10.1145/ 3170427.3188487 · DOI · Paper page
- Sara Kiesler, Jane Siegel, and Timothy W McGuire. 1984. Social psychological aspects of computer-mediated communication. American psychologist 39, 10 (1984), 1123.
- François Mairesse and Marilyn Walker. 2006. Automatic recognition of personal- ity in conversation. In Proceedings of the Human Language Technology Conference of the NAACL, Companion Volume: Short Papers. 85-88.
- D Harrison McKnight, Vivek Choudhury, and Charles Kacmar. 2002. Devel- oping and validating trust measures for e-commerce: An integrative typology. Information systems research 13, 3 (2002), 334-359.
- Gary S Nickell and John N Pinto. 1986. The computer attitude scale. Computers in human behavior 2, 4 (1986), 301-306.
- James W Pennebaker, Matthias R Mehl, and Kate G Niederhoffer. 2003. Psycho- logical aspects of natural language use: Our words, our selves. Annual review of psychology 54, 1 (2003), 547-577.
- Ronald E Robertson, Alexandra Olteanu, Fernando Diaz, Milad Shokouhi, and Peter Bailey. 2021. "I Can't Reply with That": Characterizing Problematic Email Reply Suggestions. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. 1-18.
- Julian B Rotter. 1971. Generalized expectancies for interpersonal trust. American psychologist 26, 5 (1971), 443.
- Joseph B Walther. 1996. Computer-mediated communication: Impersonal, inter- personal, and hyperpersonal interaction. Communication research 23, 1 (1996), 3-43.
- Joseph B Walther. 2011. Theories of computer-mediated communication and interpersonal relations. The handbook of interpersonal communication 4 (2011), 443-479.
- Authors
- Liu, Y., Mittal, A., Yang, D., & Bruckman, A.
- Affiliations
- Georgia Institute of Technology (US)
- Venue
- CHI Conference on Human Factors in Computing Systems
- Publisher
- ACM
- Published
- 2022-04-29
- Pages
- 1-13
- OpenAlex
- W4224979739
- Cited by
- 92
- References
- 35
- Citations by year
- 2026: 20 · 2025: 34 · 2024: 22 · 2023: 14 · 2022: 1
- FWCI
- 9.81
- Type
- conference paper
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- CHI Conference on Human Factors in Computing Systems (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › AI in Service Interactions
- Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
- Physical Sciences › Computer Science › Human-Computer Interaction › Innovative Human-Technology Interaction
- Keywords
- AI-mediated communication, large language models, trust in AI, interpersonal communication, transactional communication, AI transparency, user perceptions of AI
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.
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
