[Paper]
This data contains survey answers from 295 demographically diverse, US-based participants on AI usage. More specifically, the survey contained three following sections.
- Use Cases: Participants were asked to brainstorm use cases of AI and choose a use case that might be most drastically changed.
- Harms and Benefits: Participants to write harms and benefits of both developing and not developing the use case in free-text format, identify the groups impacted most, and rate the scale of impact.
- Application Decision: Participants were asked to make a decision on whether the application should be developed.
In data
folder
codes
: Folders of codes and their definition (code-def-0107v1.json
), fewshot examples used for prompting (fewshot-0107v1.json
), and prompt template for applying codes with an LLM (prompt-template-0107v1.json
)task-coding
: Q7benefits-coding
: Q8_4harms-coding
: Q11_4, Q11_5, Q14_4, Q14_5not-dev-benefits-coding
: Q20_4not-dev-harms-coding
: Q17_4groups-coding
: Q9, Q12, Q15, Q18, Q21
coding_results
: Coding results from LLM code applicationhuman_annotations
: Human applied coding results of 80 samplesthemes
: High level themes of each coderaw_data.csv
: Answer to all questions, including demographics and AI literacy questions
Additionally, all questions are in questions.json
. See survey.pdf
for survey format presented to participants.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.