Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 28 of the 29 references it cites were found in a registry or an index; 7 of the 15 sources its claims cite were checked for support.

Generating Together: Lessons Learned from Developing an Educational Visual Novel with AI Collaboration

Open access
Shields, Samuel, Calderwood, Alexander, Johnson-Bey, Shi, Wardrip-Fruin, Noah, Mateas, Michael, Melcer, Edward · 2024 · 2024 IEEE Conference on Games (CoG)

Abstract

Visual novels are a popular game genre for educational games. However, they often feature pre-authored plot structures that cannot dynamically adjust to the player’s progression through learning objectives. Employing procedural storytelling techniques boosts plot dynamism, but this comes at the cost of needing a larger repository of content (dialogue and images) to support different learning progressions and objectives. In this paper, we present postmortem-style case studies describing the lessons we learned from attempting to integrate large-language models (LLMs) and text-to-image models into the development of an educational visual novel about responsible conduct of research. Specifically, we discuss our experiences employing generative AI in our dialogue, character sprite, and background image creation processes.

Overview

CoChat Summary

TL;DR

This paper presents postmortem-style case studies from integrating large language models and text-to-image models into the development of an educational visual novel about responsible conduct of research. It finds generative AI useful for augmentation and brainstorming but not as a full replacement for writers and artists, requiring heavy human oversight and proving unsuitable for background image generation.

Key findings

  • ChatGPT could generate utterances meeting high-level goals, but final content needed heavy revision: text was verbose and literal, dialogue was stiff, and very little LLM-generated text made it through multiple rewrites.
  • Stable Diffusion 1.5 fine-tuned on a roughly 20-image corpus, with training sets of 15–20 images, produced usable character portraits but required postprocessing, many generation attempts due to non-determinism, and did not accelerate the overall pace of development.
  • Stable Diffusion was judged not appropriate for location backgrounds because objects were excluded or moved, spatial layout changed between images, perspective was often incorrect, and object scale varied unrealistically.
  • The production team comprised 3 writers, 1 2D/3D artist, 1 generative artist, 3 system engineers, and 1 producer; the paper frames generative techniques as creative augmentation rather than artist displacement.

Methods

  • Design: postmortem-style case studies of integrating generative AI into game production, following the postmortem approach of highlighting goals, technology, what went right, what went wrong, and what should have been done instead.
  • Data/sample: the development process of Academical 2.0, an educational visual novel about responsible conduct of research, with three case studies covering narrative generation with ChatGPT, character portrait generation with Stable Diffusion 1.5, and location background generation with Stable Diffusion.
  • Analysis: qualitative lessons learned from an interdisciplinary team of writers, artists, and programmers, including writer’s room practices, concept formation, model fine-tuning, and artist interview responses.

Limitations

  • Stated: future versions of Academical will lessen the use of generative AI in production due to the limitations encountered; the project occurred in an academic environment with low resources and contributors focused on training specific roles.
  • Evident: no comparison study between Academical 1.0 and 2.0 had yet been conducted, as it is listed as future work, so learning-outcome and ARC-factor effects are not evaluated here.
  • Evident: the findings come from a single project’s qualitative postmortem and artist interview, with no quantitative benchmark or generalizable production study reported.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/cog60054.2024.10645553

Cites

28 of 28 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

7 Valid · 1 Not valid · 8 Not checked · 37 Own results and interpretations

22 more claims

Create a free account to see all 53 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 14
  • It may state two separate claims as one. 18

Create a free account to see these claims.

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?

Citations

Cites

Show all 29 references

Research works printed but not linked: 9

  1. Martens, Procedural narrative generation
  2. Yannakakis, Mixed-initiative cocreativity
  3. Whitten, 2023, Introducing unity muse and unity sentis, ai-powered creativity, Unity Blog
  4. Burnes, 2023, Introducing nvidia ace for games-spark life into virtual characters with generative ai, NVIDIA GeForce
  5. Fowler, 2001, The agile manifesto, Software development, 9, 28
  6. Junius, 2019, Puppitor: Building an Acting Interface for Videogames. University of California
  7. Samuel, Leveraging procedural narrative and gameplay to address controversial topics
  8. Hills, 2022, Management fads: How to resist shiny object syndrome, Physician Leadership Journal, 9
  9. Dini, 2023, 7-cocreation of serious games to acquire a professional skill, EXECUTIVE COMMITTEE

Cited by

See all 11

Metadata

Authors
Shields, S., Calderwood, A., Johnson-Bey, S., Wardrip-Fruin, N., Mateas, M., & Melcer, E.
Affiliations
  • University of California, Santa Cruz (US)
  • University of California, Santa Cruz,Computational Media,Santa Cruz,United States
Venue
2024 IEEE Conference on Games (CoG)
Publisher
IEEE
Published
2024-08-05
Pages
1-8
OpenAlex
W4401943823
Cited by
11
References
29
Citations by year
2026: 10 · 2025: 1
FWCI
1.74
Type
conference paper
Language
en
Open access
Locations
  • IEEE Conference on Games (CoG) (publishedVersion)
  • eScholarship (California Digital Library) (submittedVersion, cc-by)
Funders
National Science Foundation
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence in Games
  • Social Sciences › Psychology › Developmental and Educational Psychology › Educational Games and Gamification
  • Social Sciences › Education › Artificial Intelligence in Education
Keywords
generative AI, large language models, text-to-image models, responsible conduct of research, game development, dialogue generation, collaborative game development

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Generating Together: Lessons Learned from Developing an Educational Visual Novel with AI Collaboration. https://app.cochat.ai/papers/10.1109/cog60054.2024.10645553
The paper
Shields, S., Calderwood, A., Johnson-Bey, S., Wardrip-Fruin, N., Mateas, M., & Melcer, E. (2024). Generating Together: Lessons Learned from Developing an Educational Visual Novel with AI Collaboration. 2024 IEEE Conference on Games (CoG). https://doi.org/10.1109/cog60054.2024.10645553

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.