Unmet Creativity Support Needs in Computationally Supported Creative Writing

Probably not a substantial claim: may not pass S4 (based).

Since refinement of expressive intent has only recently been recognized as an explicit goal for creativity support tools in some contexts, relatively little work has been done to provide computational support for intent refinement in storytelling contexts.

“Since refinement of expressive intent has only recently been recognized as an explicit goal for creativity support tools in some contexts, relatively little work has been done to provide computational support for intent refinement in storytelling contexts.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Distinguishing between story and discourse is one promising first step for reader experience support, since it allows a narrative generation engine to retell the same story in different ways.

“Distinguishing between story and discourse is one promising first step for reader experience support, since it allows a narrative generation engine to retell the same story (plot-wise) in different ways (Rishes et al., 2013).”

• cites Rishes et al., 2013Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

As stories and their casts of characters grow in size and more of the fictional timeline is filled in, the author runs increasing risk of introducing inconsistencies, plot holes, or unexplained situations that may break the reader's ability to suspend disbelief.

“As stories and their casts of characters grow in size, and more of the fictional timeline is filled in, the author runs increasing risk of introducing inconsistencies (conflicting factual assertions or implications), plot holes, or unexplained situations that may break the reader's ability to suspend disbelief.”

Probably not a substantial claim: may not pass S4 (based).

Language model continuations of human-written text tend to be syntactically valid and relevant to storyworld entities or situations described in the immediately preceding text, enabling them to function as viable short-term suggestions.

“Language model continuations of humanwritten text tend to be syntactically valid and relevant to storyworld entities or situations that were described in the immediately preceding text, enabling them to function as viable short-term suggestions for what might happen next in a written story.”

May combine two claims.

Machine-learned language models are good at capturing local coherence, but tend to struggle with the global constraints implied by plot structure.

“Machine-learned language models are good at capturing local coherence, but tend to struggle with the global constraints implied by plot structure.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The key technological tool for maintaining consistency is a world model, or a computational representation of the diegetic phenomena that a story aims to fictionalize.

“The key technological tool for maintaining consistency is a world model, or a computational representation of the diegetic phenomena that a story aims to fictionalize.”

Probably not a substantial claim: may not pass S3 (load-bearing).

If story authors strategically withhold information, they can elicit inferences on the part of readers to fill in the gaps in ways that can evoke humor, shock, or horror understanding.

“If story authors strategically withhold information, they can elicit inferences on the part of readers to fill in the gaps in ways that can evoke humor, shock, or horror understanding (Cohn, 2019).”

• cites Cohn, 2019Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

By representing a world model in its own right, one can specify consistency constraints as first-order logic formulas whose constituent predicates refer to the world model.

“By representing a world model in its own right, one can specify consistency constraints as (e.g.) first-order logic formulas whose constituent predicates refer to the world model.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

To support authors in crafting an intentional experience for their readers, computational tools need to be able to reason about or simulate the reader's cognitive processes.

“To support authors in crafting an intentional experience for their readers, computational tools need to be able to reason about (or perhaps even simulate) the reader's cognitive processes.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

One difficulty in creative work is that the creator may not know exactly what they are trying to express, and the expressive intent may shift as the creator's understanding of the work evolves.

“One difficulty in creative work is that the creator themselves may not know exactly what they are trying to express, and the expressive intent may shift as the creator's understanding of the work evolves.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Poor pacing can cause a reader to get bored or overwhelmed with the story and fail to connect with the characters or the underlying message.

“Poor pacing can cause a reader to get bored or overwhelmed with the story and fail to connect with the characters or the underlying message that the writer is attempting to convey.”

May combine two claims.

Creative writers face additional challenges, including maintaining narrative consistency, developing plot structure, architecting reader experience, and refining their expressive intent.

“However, creative writers face additional challenges, including maintaining narrative consistency, developing plot structure, architecting reader experience, and refining their expressive intent”

• Also stated in: § 1 Introduction · boosted

  “have revealed a preliminary inventory of several distinct but interrelated creativity support needs among creative writers, including:”

• cites Kreminski et al., 2019bNot checked yet

  Not checked in the free run.

• cites Kreminski and Wardrip-Fruin, 2019Not checked yet

  Not checked in the free run.

• cites Kreminski et al., 2020aNot checked yet

  Not checked in the free run.

May combine two claims.

Current large language models do not directly address consistency maintenance, longer-term plot structure, management of reader experience, or the challenge of refining high-level expressive intent.

“However, they do not directly address consistency maintenance, longer-term plot structure, management of reader experience, or the challenge of refining high-level expressive intent”

• Also stated in: § Abstract · plain

  “which are not well-addressed by current LLM-backed tools”

May combine two claims.

The five creative writing support needs are getting unstuck, maintaining consistency, constructing a satisfying overall story arc including a conclusion/resolution, managing reader experience, and refining and iterating on expressive intent.

“• Getting unstuck • Maintaining consistency • Constructing a satisfying overall story arc, including a conclusion/resolution”

• Also stated in: § 1 Introduction · plain

  “• Managing reader experience”

• Also stated in: § 1 Introduction · plain

  “• Refining and iterating on expressive intent”

• cites Kreminski et al., 2019bNot checked yet

  Not checked in the free run.

• cites Kreminski and Wardrip-Fruin, 2019Not checked yet

  Not checked in the free run.

• cites Kreminski et al., 2020aNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Reasoning about reader experience requires a good understanding of how stories work at a cognitive level.

“Reasoning about reader experience requires a good understanding of how stories work at a cognitive level: e.g., that readers work as problem solvers when processing narrative text, working to stay one step ahead of the story to make sense of what has happened so far and predict what will happen next (Gerrig and Bernardo, 1994).”

• cites Gerrig and Bernardo, 1994Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

A written work has an experiential impact on its readers, and understanding the levers that affect that impact is a key part of narrative intelligence.

“Likewise, a written work has an experiential impact on its readers, and understanding the levers that affect that impact is a key part of narrative intelligence.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Plot structures describe global rather than local features of a text, and they have more to do with the underlying world model than with the specific actions or events inferable from lexical properties of the text.

“Importantly, plot structures describe global rather than local features of a text, and they have more to do with the underlying world model (see previous section) than they do with the specific actions or events that are inferable from lexical properties of the text.”

Probably not a substantial claim: may not pass S2 (checkable).

The ability to directly author and manipulate knowledge corresponding to a situation model is central to a fiction author's task.

“The ability to directly author and manipulate knowledge corresponding to a situation model (or similar) is central to a fiction author's task.”