Disclosing Generative AI Use in Digital Humanities Research
Graphs/Analysis
Project Developers
Menu
- Overview
- Note on Viewing This Page
- Do Scholars Think AI Use Should Be Disclosed?
- Who Should Be Responsible for Ensuring AI Disclosure?
- Does Context Matter for AI Disclosure?
- Which AI-Assisted Activities Do Scholars Believe Should Be Disclosed?
- Where Do AI Disclosure Expectations Align with Practice?
- Do AI Disclosure Principles Overall Align with Disclosure Practices?
- Bonus: De-identified Survey Responses
Overview
This project investigates how digital humanists perceive and approach generative AI (GenAI) disclosure in research. It is based on a survey that we conducted in 2025. The interactive graphs that follow offer a series of visual representations based on our data.
For each graph, we include the survey question on which the graph is based, the graph itself, guidance on reading the graph, and our analysis of the graph. There is also a "bonus" section that offers access to a de-identified version of the survey responses for our project. Further information about the project, participants, and methodology appears on our home page.
Note on Viewing This Page
This page includes a series of graphs. Site visitors can view and filter the graphs on mobile devices, but desktop browsers offer the best option for engaging with the graphs given the complexity of their data. On desktop, hovering over a graph shows additional detail. On touch devices, tapping a graph element shows the same detail in a panel below the graph.
Every graph is also accompanied by a data table. Screen readers read these tables in place. Below each graph, a “Show data table on screen” button displays the table visually, so all of the data can be reached without a mouse.
Do Scholars Think AI Use Should Be Disclosed?
Survey responses on the necessity of disclosing GenAI use (all respondents, n=107)
| Response | Count | Percentage |
|---|
How to Read This Graph
This graph shows how respondents assess the necessity of disclosing generative AI use in scholarship. Each bar represents a response option. Options range from "very necessary" to "not necessary at all". The height of each bar indicates how many respondents selected that option.
The filters above allow users to explore how views on disclosure vary by categories like "experience," "seniority," "discipline," or self-defined "genAI literacy." When a demographic category is selected, a second filter appears to choose a specific value. The chart then updates to show responses only from that group. Values reflect the number of respondents who selected a given answer. On desktop view, users can also hover over a given bar to see percentages of respondents who selected a given option.
Graph Analysis
Coming soon!
Who Should Be Responsible for Ensuring AI Disclosure?
Communities
Survey responses on which entities should be responsible for ensuring GenAI disclosure (all respondents, n=107)
| Entity | Selections |
|---|---|
| Journal publishers and editorial boards | 92 |
| Academic institutions | 91 |
| Professional organizations | 73 |
| Funding bodies | 72 |
| Ethical review boards | 63 |
| Government agencies | 54 |
| Industry organizations | 38 |
| Independent oversight bodies | 38 |
| International organizations | 37 |
| Other | 7 |
| Prefer not to answer | 5 |
Survey responses on how often pairs of entities were selected together, from most to least often (all respondents, n=107)
| Entity pair | Co-selections |
|---|---|
| Academic institutions ↔ Journal publishers and editorial boards | 84 |
| Academic institutions ↔ Professional organizations | 73 |
| Academic institutions ↔ Funding bodies | 70 |
| Professional organizations ↔ Journal publishers and editorial boards | 70 |
| Funding bodies ↔ Journal publishers and editorial boards | 70 |
| Academic institutions ↔ Ethical review boards | 60 |
| Journal publishers and editorial boards ↔ Ethical review boards | 59 |
| Professional organizations ↔ Funding bodies | 58 |
| Professional organizations ↔ Ethical review boards | 56 |
| Government agencies ↔ Academic institutions | 54 |
| Funding bodies ↔ Ethical review boards | 54 |
| Government agencies ↔ Journal publishers and editorial boards | 53 |
| Government agencies ↔ Professional organizations | 48 |
| Government agencies ↔ Funding bodies | 48 |
| Government agencies ↔ Ethical review boards | 44 |
| Academic institutions ↔ International organizations | 37 |
| Journal publishers and editorial boards ↔ Independent oversight bodies | 37 |
| Academic institutions ↔ Industry organizations | 36 |
| Academic institutions ↔ Independent oversight bodies | 36 |
| Professional organizations ↔ Industry organizations | 36 |
| Journal publishers and editorial boards ↔ International organizations | 36 |
| Professional organizations ↔ Independent oversight bodies | 35 |
| Professional organizations ↔ International organizations | 35 |
| Journal publishers and editorial boards ↔ Industry organizations | 35 |
| Funding bodies ↔ Independent oversight bodies | 34 |
| Funding bodies ↔ International organizations | 34 |
| Ethical review boards ↔ Independent oversight bodies | 34 |
| Ethical review boards ↔ International organizations | 34 |
| Government agencies ↔ Independent oversight bodies | 33 |
| Government agencies ↔ International organizations | 33 |
| Funding bodies ↔ Industry organizations | 33 |
| Industry organizations ↔ Ethical review boards | 33 |
| Government agencies ↔ Industry organizations | 32 |
| Industry organizations ↔ Independent oversight bodies | 30 |
| Industry organizations ↔ International organizations | 29 |
| Independent oversight bodies ↔ International organizations | 28 |
| Academic institutions ↔ Other | 5 |
| Journal publishers and editorial boards ↔ Other | 5 |
| Government agencies ↔ Other | 4 |
| Professional organizations ↔ Other | 4 |
| Funding bodies ↔ Other | 4 |
| Industry organizations ↔ Other | 3 |
| Ethical review boards ↔ Other | 3 |
| International organizations ↔ Other | 3 |
| Independent oversight bodies ↔ Other | 2 |
How to Read This Graph
This network graph shows how respondents understand responsibility for generative AI disclosure as a shared task. Each circle represents a given entity (e.g., "academic institutions," "funders," or "publishers"). Larger circles indicate entities selected by more respondents.
Lines between circles show co-selection, i.e., entities that respondents often chose together. Thicker lines indicate stronger connections, while thinner lines indicate weaker connections.
On desktop view, users can hover over circles or connections to see detailed counts.
Graph Analysis
Coming soon!
Does Context Matter for AI Disclosure?
Survey responses on whether scholarly product type matters for GenAI disclosure (all respondents, n=81)
| Response | Count | Percentage |
|---|
Survey responses on which scholarly products require GenAI disclosure, by response to Q13 (all respondents, n=107)
| Scholarly Product | Always matters | Usually matters | Sometimes matters | Rarely matters | Never matters | Prefer not to answer | Total |
|---|
How to Read This Graph
These graphs combine responses to Q13 and Q14 to show how views on genAI disclosure vary by scholarly product type. The two graphs demonstrate that context matters for AI disclosure, and within each context scholars’ expectations vary.
The pie chart (Q13) shows whether respondents believe the type of scholarly product has a bearing on the need for disclosure.
The stacked bar chart (Q14) shows which scholarly products respondents believe to require genAI disclosure. Each bar represents a product type, and its total length reflects how many respondents selected it.
Users can apply the filters above to explore how patterns vary by categories like "experience," "seniority," "discipline," or self-defined "genAI literacy." When a demographic category is selected, a second filter appears to choose a specific value.
Graph Analysis
Coming soon!
Which AI-Assisted Activities Do Scholars Believe Should Be Disclosed?
Survey responses on which activities should be disclosed (all respondents, n=107)
| Activity | Count |
|---|---|
| Writing and rewriting | 62 |
| Data collection, transcription, translation, and analysis | 44 |
| Peer review | 35 |
| Create graphics, pictures, videos or audios | 32 |
| Summarizing, explaining, and simplifying content | 25 |
| Literature search and review | 24 |
| Brainstorming, including creating hypotheses or arguments | 21 |
| Data organization, representation, and visualization | 21 |
| Critiquing and offering alternate perspectives | 11 |
| Coding and Programming | 9 |
| Other (please specify) | 6 |
| Project design and planning | 5 |
| Dissemination (e.g., website building, conference presentation, social media marketing) | 4 |
| Prefer not to answer | 1 |
| Topic or skill coaching | 0 |
How to Read This Graph
This graph shows which generative AI-assisted research activities respondents believe are most important to disclose. Each bar represents one activity, and its height reflects how many respondents selected it. Taller bars suggest stronger consensus that an activity warrants disclosure, while shorter bars indicate less consensus.
Users can use the filters above to explore how disclosure priorities differ by categories like "experience," "seniority," "discipline," or self-defined "genAI literacy." When a demographic category is selected, a second filter appears to choose a specific value. Counts show raw numbers of respondents per item.
Graph Analysis
Coming soon!
Where Do AI Disclosure Expectations Align with Practice?
Survey responses on expected disclosure (Q15) and actual disclosure practice (Q21), by activity (all respondents, n=107)
How to Read This Graph
This heatmap graph compares perceptions of what should be disclosed (Q15) with actual genAI use (Q21).
- Rows (Q15) show activities respondents believe are important to disclose.
- Columns (Q21) show activities for which respondents report using generative AI.
- Each cell shows how many respondents report using genAI and disclosing such use for a given activity.
Darker cells indicate stronger alignment between practice and disclosure, while lighter cells indicate weaker alignment. Because both questions (Q15 and Q21) allow for multiple selections, the heatmap graph also highlights broader patterns.
Use the checkboxes to focus on specific activities and explore where norms of disclosure closely align or diverge from actual genAI use.
Graph Analysis
Coming soon!
Do AI Disclosure Principles Overall Align with Disclosure Practices?
Survey responses on the necessity of disclosure (Q6) and the degree of disclosure (Q16) (all respondents, n=107)
| Necessity of disclosure (Q6) | Degree of disclosure (Q16) | Count |
|---|---|---|
| Very necessary | Minimal disclosure | 1 |
| Very necessary | Brief disclosure | 33 |
| Very necessary | Detailed disclosure | 33 |
| Somewhat necessary | Minimal disclosure | 2 |
| Somewhat necessary | Brief disclosure | 18 |
| Somewhat necessary | Detailed disclosure | 1 |
| Neutral | Minimal disclosure | 4 |
| Neutral | Brief disclosure | 4 |
| Neutral | Detailed disclosure | 1 |
| Neutral | Prefer not to answer | 1 |
| Not very necessary | Minimal disclosure | 1 |
| Not very necessary | Brief disclosure | 3 |
| Not very necessary | Prefer not to answer | 1 |
| Not necessary at all | Brief disclosure | 1 |
| Not necessary at all | Detailed disclosure | 2 |
| Prefer not to answer | Prefer not to answer | 1 |
Survey responses on the degree of disclosure (Q16) and the frequency of disclosure (Q10) (all respondents, n=107)
| Degree of disclosure (Q16) | Frequency of disclosure (Q10) | Count |
|---|---|---|
| Minimal disclosure | Never | 5 |
| Minimal disclosure | 1-2 times | 1 |
| Minimal disclosure | More than 10 times | 1 |
| Minimal disclosure | Prefer not to answer | 1 |
| Brief disclosure | Never | 36 |
| Brief disclosure | 1-2 times | 17 |
| Brief disclosure | 3-5 times | 3 |
| Brief disclosure | 6-10 times | 1 |
| Brief disclosure | More than 10 times | 1 |
| Brief disclosure | Prefer not to answer | 1 |
| Detailed disclosure | Never | 31 |
| Detailed disclosure | 1-2 times | 4 |
| Detailed disclosure | 3-5 times | 1 |
| Detailed disclosure | 6-10 times | 1 |
| Prefer not to answer | Never | 1 |
| Prefer not to answer | 3-5 times | 2 |
Survey responses on the frequency of disclosure (Q10) and whether disclosure was required or voluntary (Q11) (all respondents, n=107)
| Frequency of disclosure (Q10) | Required or voluntary (Q11) | Count |
|---|---|---|
| Never | Never disclosed | 73 |
| 1-2 times | It was required | 7 |
| 1-2 times | Both required and voluntary | 1 |
| 1-2 times | I disclosed it voluntarily | 14 |
| 3-5 times | It was required | 2 |
| 3-5 times | I disclosed it voluntarily | 4 |
| 6-10 times | I disclosed it voluntarily | 2 |
| More than 10 times | Both required and voluntary | 1 |
| More than 10 times | Other | 1 |
| Prefer not to answer | I disclosed it voluntarily | 1 |
| Prefer not to answer | Prefer not to answer | 1 |
How to Read This Graph
This Sankey graph shows how respondents’ answers connect across different survey questions and so reveals how views, practices, and decisions tend to cluster or diverge.
Each block represents a possible answer, and the flows between blocks show how many respondents chose both of those answers. Wider flows mean more respondents followed that same pathway.
These flows make it easy to see where respondents’ perspectives and practices align, diverge, or take different directions rather than following a single pattern.
Graph Analysis
Coming soon!
Bonus: De-identified Survey Responses
Download the original de-identified survey responses and a question-by-question visualization of these responses.