Improving Human-AI Interactions in the Future of Work (Experimental Systems / Game Building)
Park Sinchaisri, Professor
Business, Haas School
Open. Apprentices for the Fall semester needed. Enter your application online beginning August 21st. You must also provide a PDF of your academic summary as part of applying to this mentor. Instructions will be included in the confirmation email after you submit your regular application. The deadline to apply is Monday, August 31st, 4pm.
BOBALAB designs interactive experiments to study how humans learn, make decisions, and collaborate with algorithms and AI.
Many of our research questions cannot be answered from observational data alone. We therefore build online environments where participants make repeated decisions, receive algorithmic recommendations, learn from feedback, interact with AI or other humans, and face incentives based on their choices.
Current and planned projects include sequential decision games involving precise versus broad AI advice; human-AI delegation tasks in which people choose whether to act, ask, or defer; experiments on learning to use unfamiliar AI tools; performance-feedback and algorithmic-management studies; and multi-agent coordination environments.
Our work has been published in Management Science, M&SOM, and PLOS ONE, and presented at leading CS/HCI conferences including CHI, ICML, and CSCW. BOBALAB alumni have gone on to graduate programs at Harvard, MIT, Stanford, Berkeley, Penn, CMU, Northwestern, Michigan, and other institutions.
Role: Students in this track work at the intersection of research design, product design, and software engineering. They do not simply receive a finished specification and build a webpage. Good experiments require thinking carefully about participant understanding, incentives, experimental validity, data collection, and all of the ways a study might fail.
Typical tasks include:
- Translating research ideas into intuitive interactive experiences
- Building web-based experiments and games
- Implementing randomization, experimental conditions, incentives, timers, state transitions, recommendations, and feedback
- Connecting experimental interfaces to backend services, models, APIs, or databases
- Designing reliable event-level logging and data schemas
- Building automated tests and simulation bots for experimental environments
- Piloting studies and identifying confusing instructions or unintended strategies
- Debugging edge cases and hardening experiments before deployment
- Producing clear documentation and data dictionaries
Students will learn how behavioral experiments are designed from the ground up, how research questions translate into software, how to instrument systems for reliable scientific measurement, and how to balance experimental validity with participant experience.
Qualifications: - Strong JavaScript/TypeScript or comparable web-development skills
- Evidence of being able to build things independently
- Excellent debugging skills and unusually strong attention to detail
- Comfort working with incomplete specifications and proposing sensible solutions
- Strong testing mindset and willingness to chase edge cases
- Ability to document design and implementation decisions clearly
- Bonus: React, Node, Firebase/Supabase, jsPsych, oTree, Qualtrics JavaScript, APIs, databases, or cloud deployment
We especially like applicants who have built something they cared about: a game, app, website, research tool, hackathon project, or unusual side project.
AI coding tools are encouraged, but “the AI wrote it” is not an explanation for how an experimental system works. A subtle bug can invalidate an entire study. Students must understand important logic, inspect generated code, test aggressively, and take responsibility for reliability.
Hours: 9-11 hrs
Off-Campus Research Site: Remote is possible.
Related website: https://parksinchaisri.github.io/files/paper-tips.pdf
Related website: https://parksinchaisri.github.io/files/paper-tips.pdf