Microsoft · 01 / 05A year of Copilot studies became a trust framework for when AI should ask, suggest, or act.
Hi, I’m Priyal.
I research how people and technology get along, and where they don’t.
Four years of turning that research into product and policy decisions. Lately, on AI.
I use mixed-method research to help teams build trust in technology across cultures and literacy levels.
Currently at Carnegie Mellon · Previously Microsoft, Razorpay & Meesho
Right now
- Using LLMs as a data collection methodology
- Studying effects of AI manipulation disclosure
- Building framework for Responsible AI adoption through NSF AI-SDM
Projects
Selected work
- Industry research
-
How can Copilot earn users’ trust?
- Research synthesis
- Literature review
- Social listening
- Question
- What do people need before letting Copilot act for them?
- Evidence
- Findings from a year of Copilot studies.
- Decision
- Give people clear information and control before asking them to trust more automation.
-
What makes people leave Outlook on mobile web?
- Interviews
- Survey
- Thematic analysis
- Question
- Why do mobile-web users stay or leave?
- Evidence
- 36 interviews and more than 300 survey responses from six types of users.
- Decision
- Improve how people move between their accounts and email apps.
-
How can we optimize onboarding for merchants with low digital literacy?
- Session review
- Ticket analysis
- Usability testing
- Question
- Why were small merchants leaving the identity-verification form?
- Evidence
- Support tickets, recordings of people using the form, and role-playing exercises.
- Decision
- Ask one plain-language question at a time and explain why.
-
How can research keep improving a website right up to launch?
- Pre-mortem
- UX audit
- Impression test
- Perception test
- Usability test
- Preference test
- Question
- What would the redesigned homepage communicate before launch?
- Evidence
- Six kinds of research completed over six weeks.
- Decision
- Fix the main problems before the new site launched.
-
Should Razorpay build a consumer rewards product?
- Interviews
- Focus groups
- Participatory design
- Survey
- Question
- Should Razorpay build a consumer rewards product?
- Evidence
- 40+ interviews and a survey of more than 1,000 people.
- Decision
- Replace the original idea with the rewards model people preferred.
- Publications and features
-
CHI ’26 Lost in Transcription: Subtitle Errors in Automatic Speech Recognition Reduce Speaker and Content Evaluations
Preregistered experiment · 207 participants
1ST PLACE The role of HCI in policy
Election deepfake policy proposal
UXR INDIA Tracing the history of UXR in India
Interviews and historical research
- Playground
-
Things I make for fun
Small things I make for fun
Prototypes, games, music, and other things I wanted to try.
- Experiments
- Games
About me
I make sense of messy human problems.
I started as a product designer, then moved into research because I wanted to understand the problem before designing a solution. Over the past four years, I’ve worked at Microsoft, Razorpay, and Meesho.
I plan and run interviews, surveys, usability studies, and workshops. I’ve worked with low-tech-literacy users, people across cultures and languages, and teams building trust in AI.
I’m currently completing a master’s in Human-Computer Interaction at Carnegie Mellon. My recent academic work looks at how AI changes people’s opinions, how errors in automatic captions affect speakers, and how people decide when to use AI.