# I want to understand, with a real human survey, wh... ## Conversation ## User *2026-07-02T12:28:40.874272+00:00* I want to understand, with a real human survey, what people are most hoping AI developments focus on in terms of impact e.g., AI for climate change, AI for medicine, etc. --- ## Assistant *2026-07-02T12:28:55.176155+00:00* Great question — and a timely one. Let me ask a few things before we design this, since the details will shape what we build. You mentioned wanting a **real human survey**, which is great — we can deploy a survey that people take through a web link. A couple of things I'd like to understand first: **Tool calls:** - `Skill` - `skill`: workflow-session-start --- ## User *2026-07-02T12:29:13.949541+00:00* Q: Who are you hoping to reach with this survey? For example: general public, tech workers, policymakers, your own audience/network, a specific demographic? This shapes how we frame the questions and what we learn. A: General public Q: What will you use the findings for? This helps me calibrate the depth and rigor of the survey. A: Personal curiosity --- ## Assistant *2026-07-02T12:29:33.690461+00:00* Good — general public, personal curiosity. That gives us a lot of freedom to keep it short and engaging (which also helps completion rates). One more thing that will shape the design: --- ## User *2026-07-02T12:31:55.017402+00:00* Q: How do you want to learn about people's priorities? This affects how long the survey is and what kind of findings we get. A: Deeper exploration --- ## Assistant *2026-07-02T12:32:02.300748+00:00* Great — a deeper exploration will give us richer findings. Let me read the relevant design references and put together a plan. **Tool calls:** - `Skill` - `skill`: workflow-file-layout - `Skill` - `skill`: workflow-completion-checklist - `Skill` - `skill`: edsl-quirks - `Skill` - `skill`: edsl-human-survey - `Skill` - `skill`: research-survey - `Skill` - `skill`: workflow-plan-template - `Skill` - `skill`: workflow-model-selection - `Skill` - `skill`: report-style - `Skill` - `skill`: report-transparency - `Skill` - `skill`: report-plot-standards - `Glob` - `pattern`: sessions/*/chat_summary.md Now I have all the context I need. Let me check the filesystem and then draft a plan for you. **Tool calls:** - `Bash` - `description`: Check if sessions directory exists, `command`: ls sessions/ 2>/dev/null || echo "No sessions directory yet" - `Skill` - `skill`: edsl-surveys Good, this is a fresh workspace. Let me now scaffold the study directory and write the plan for your review. **Tool calls:** - `Bash` - `description`: Scaffold study directory structure, `command`: mkdir -p sessions && python ~/.claude/skills/workflow-file-layout/scripts/create_study_project.py sessions/topic_ai-development-priorities/study_a - `Bash` - `description`: Verify study structure, `command`: ls sessions/topic_ai-development-priorities/study_a/ Now let me write the plan: **Tool calls:** - `Write` - `file_path`: /home/user/sessions/topic_ai-development-priorities/study_a/plan.md Now let me present this to you: **Tool calls:** - `Skill` - `args`: sessions/topic_ai-development-priorities/study_a/plan.md, `skill`: present-agent-work - `Bash` - `description`: Present plan.md to user, `command`: python ~/.claude/skills/present-agent-work/scripts/present_file.py ~/sessions/topic_ai-development-priorities/study_a/plan.md --title "AI Development Priorit... Here's the plan for your survey. The key design choices: - **Starts with an open-ended question** before showing any list, so you capture what's genuinely top-of-mind for people - **Then structured priorities** -- pick top 3, then force a single pick, which reveals real trade-offs - **An "aspiration gap" comparison** -- what people *hope* AI focuses on vs. where they think it's *currently* having impact. That gap is often the most interesting finding - **Three demographics** (age, education, AI familiarity) for breakdowns without making the survey too long The five decisions I'd like your input on: --- ## User *2026-07-02T12:38:39.970820+00:00* Q: How will respondents find the survey? This is the biggest design choice. A: Prolific (paid panel) Q: How many responses are you hoping for? This mainly matters if you want demographic breakdowns. A: 20-30 (directional) Q: Should I keep Q7 ('What concerns you most about how AI is being developed?')? It broadens the survey beyond hopes into worries. A: Keep it Q: Anything to change about the survey content? (Areas list, demographics, question wording, etc.) A: Looks good as-is --- ## Assistant *2026-07-02T12:39:10.825663+00:00* Great choices. Let me build this out. Here's what I'll do: 1. Build and preview the survey 2. Deploy it to Expected Parrot (get your shareable link) 3. Set up Prolific recruitment (after confirming the cost with you) 4. Once responses come in, I'll analyze and write the report **Tool calls:** - `TodoWrite` - `todos`: [{"activeForm": "Building the EDSL survey", "content": "Build the EDSL survey (10 questions)", "status": "in_progress"}, {"activeForm": "Deploying survey via... - `Write` - `file_path`: /home/user/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/build_survey.py - `Write` - `file_path`: /home/user/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/deploy_survey.py Now let me build and deploy the survey: **Tool calls:** - `Bash` - `description`: Build the survey and save it, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && python edsl_jobs/job_a/build_survey.py Survey built. Now let me deploy it so you get a live link: **Tool calls:** - `Bash` - `description`: Deploy survey via humanize, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && python edsl_jobs/job_a/deploy_survey.py - `TodoWrite` - `todos`: [{"activeForm": "Building the EDSL survey", "content": "Build the EDSL survey (10 questions)", "status": "completed"}, {"activeForm": "Deploying survey via h... The survey is live! Here are your links: - **Take the survey**: https://www.expectedparrot.com/respond/human-surveys/f97ac46b-5fa7-449c-99f7-e18a2eb7b10a - **Admin dashboard** (monitor responses): https://www.expectedparrot.com/home/human-surveys/f97ac46b-5fa7-449c-99f7-e18a2eb7b10a You can try it yourself first to make sure everything looks right. Now for Prolific recruitment -- before I launch it, here's the cost: --- ## User *2026-07-02T12:47:47.503736+00:00* Q: Ready to launch Prolific recruitment? Here's the breakdown: 25 respondents at $1.50 each = $37.50 in participant payments (plus Prolific's service fee, typically ~33%, so roughly $50 total). The $1.50/person rate works out to ~$15/hour for a 6-minute survey, which is above Prolific's minimum and fair. Want to proceed? A: Launch it (25 respondents) --- ## Assistant *2026-07-02T12:47:56.862662+00:00* Let me launch the Prolific study now: **Tool calls:** - `Bash` - `description`: Create Prolific study in draft mode, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && python -c " import json from edsl import Coop coop = Coop() # Load deployment info with ope... Prolific study created. Total cost: **$53.57** (includes Prolific's service fee). Now let me publish it to make it live: **Tool calls:** - `Bash` - `description`: Publish Prolific study to make it live, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && python -c " import json from edsl import Coop coop = Coop() with open('data/cooked/deploy_i... - `Bash` - `description`: Publish Prolific study with study_id, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && python -c " import json from edsl import Coop coop = Coop() with open('data/cooked/deploy_i... - `TodoWrite` - `todos`: [{"activeForm": "Building the EDSL survey", "content": "Build the EDSL survey (10 questions)", "status": "completed"}, {"activeForm": "Deploying survey via h... Let me also commit what we have so far: **Tool calls:** - `Bash` - `description`: Stage all files and check status, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && git add -A && git status - `Bash` - `description`: Create initial commit, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && git commit -m "$(cat <<'EOF' Initial setup: AI development priorities human survey Built an... - `Bash` - `description`: Configure git user and commit, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && git config user.email "researcher@expectedparrot.com" && git config user.name "EP Researche... Your survey is live and recruiting. Here's the summary: **Survey**: "What Should AI Focus On?" -- 10 questions, ~6 minutes **Respondent link**: https://www.expectedparrot.com/respond/human-surveys/f97ac46b-5fa7-449c-99f7-e18a2eb7b10a **Admin dashboard**: https://www.expectedparrot.com/home/human-surveys/f97ac46b-5fa7-449c-99f7-e18a2eb7b10a **Prolific**: Live, recruiting 25 respondents at $1.50 each ($53.57 total including fees) Responses typically start arriving within an hour on Prolific. You can monitor progress on the admin dashboard. **Next step**: Once responses come in (Prolific studies with 25 people usually fill within a few hours), come back and I'll pull the results, run the analysis, and write up the report with charts showing priority rankings, the aspiration gap, optimism levels, and demographic breakdowns. Just let me know when you'd like me to check for responses! --- ## User *2026-07-02T13:00:13.340696+00:00* how's it going - any responses so far? --- ## Assistant *2026-07-02T13:00:37.202590+00:00* 20 responses already! That's most of the 25 we recruited. Want me to wait for the remaining 5 to trickle in, or should I pull what we have now and start the analysis? 20 is plenty for interesting findings. **Tool calls:** - `Bash` - `description`: Check response count on deployed survey, `command`: cd /home/user/sessions/topic_ai-development-priorities/study_a && python -c " import json from edsl import Coop coop = Coop() with open('data/cooked/deploy_i... --- ## User *2026-07-02T13:02:47.508861+00:00* Q: We have 20 of 25 responses. How would you like to proceed? A: Analyze now with 20 --- ## Assistant *2026-07-02T13:02:54.766643+00:00* Let me pull the results and get to work. 21 responses in. Let me explore the data: Fascinating data. Healthcare is the clear winner -- 43% of respondents chose it as *the* single area AI should focus on. And there's a striking aspiration gap: healthcare is the top hope but education is where people actually see AI having impact now. Let me build the analysis and report. Now let me build the analysis scripts. I'll create them in parallel: Now let me run all the analysis scripts: Let me check the plots look right: All four plots look clean. Now let me push the study objects to Expected Parrot and write the report. Now let me write the report. First, let me generate the survey appendix: Now let me compile to PDF and HTML: Let me verify the PDF looks right: PDF looks great -- clean layout, charts rendering nicely. Let me check the remaining pages and then wrap up. The full report looks great -- 8 pages, all charts clean, quotes properly formatted, appendix table readable. Let me commit and present the final deliverable. Now let me present the final report: Here's your completed report. The three headline findings: 1. **Healthcare dominates.** 43% chose it as *the* single priority -- 3x more than anything else. Even unprompted (before seeing any list), most people spontaneously mentioned medicine/health. 2. **The aspiration gap is striking.** The biggest disconnect: 57% want AI to focus on healthcare but only 10% see it working there now (+48 pp gap). Climate has a 38-0 gap -- everyone wants it, nobody sees it happening. Education is the one reversal: more people see AI working there (38%) than wish it would (33%). 3. **Cautious optimism.** Mean 5.1 on a 7-point scale -- people lean positive but aren't starry-eyed. 76% scored 5+, but only 14% hit the maximum. **Your deliverables:** - [PDF report](writeup/report.pdf) (8 pages) and [HTML version](writeup/report.html) - [Survey on Expected Parrot](https://www.expectedparrot.com/content/65114144-393c-4c0f-aa0f-543dbf611394) - [Full results (21 responses)](https://www.expectedparrot.com/content/73015f60-5c3c-45a2-8615-c250349adbfa) The remaining 4 Prolific responses may still trickle in -- let me know if you'd like me to re-pull and update the report once they're all in. Otherwise, this is complete! --- ## Files - [**report.html**](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/report.html) (469.3 KB) - [Dockerfile](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/Dockerfile) (0 B) - [Makefile](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/Makefile) (0 B) - [README.md](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/README.md) (0 B) - [plan.md](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/plan.md) (8.3 KB) - [requirements.txt](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/requirements.txt) (0 B) - [plot_aspiration_gap.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/analysis/plot_aspiration_gap.py) (5.3 KB) - [plot_optimism.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/analysis/plot_optimism.py) (2.3 KB) - [plot_single_priority.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/analysis/plot_single_priority.py) (2.5 KB) - [plot_style.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/analysis/plot_style.py) (12.1 KB) - [plot_top_three.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/analysis/plot_top_three.py) (2.9 KB) - [table_summary.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/analysis/table_summary.py) (4.6 KB) - [survey.json.gz](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/computed_objects/survey.json.gz) (1.9 KB) - [results.json.gz](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/data/results.json.gz) (15.8 KB) - [appendix.md](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/appendix.md) (0 B) - [appendix_survey.md](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/appendix_survey.md) (3.9 KB) - [ep_logo.pdf](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/ep_logo.pdf) (4.6 KB) - [fix-table-widths.lua](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/fix-table-widths.lua) (4.8 KB) - [planning.md](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/planning.md) (0 B) - [report.css](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/report.css) (4.7 KB) - [report.md](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/report.md) (13.0 KB) - [report.pdf](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/report.pdf) (394.1 KB) - [report_header.tex](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/report_header.tex) (2.2 KB) - [deploy_info.json](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/data/cooked/deploy_info.json) (327 B) - [prolific_info.json](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/data/cooked/prolific_info.json) (858 B) - [build_survey.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/build_survey.py) (6.0 KB) - [deploy_survey.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/deploy_survey.py) (2.5 KB) - [run_job.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/run_job.py) (0 B) - [study_agent_list.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/study_agent_list.py) (0 B) - [study_model_list.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/study_model_list.py) (0 B) - [study_scenario_list.py](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/edsl_jobs/job_a/study_scenario_list.py) (0 B) - [aspiration_gap.png](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/plots/aspiration_gap.png) (112.0 KB) - [optimism.png](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/plots/optimism.png) (66.1 KB) - [single_priority.png](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/plots/single_priority.png) (74.2 KB) - [top_three.png](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/plots/top_three.png) (84.2 KB) - [demographics.csv](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/tables/demographics.csv) (465 B) - [priorities.csv](https://www.expectedparrot.com/api/macaw/share/uUlwjBqjSFO-AnV_TD4CPw/files/sessions/topic_ai-development-priorities/study_a/writeup/tables/priorities.csv) (391 B)