Surveys

What are the tradeoffs between detecting adversaries and protecting privacy in survey design?

Backstory!

Early in my PhD, I designed a table-based programming language and runtime system to design, debug, and deploy scientific surveys on the web. With my project adivsor, I collaborated with folks from the Linguistics department. I first presented SurveyMan at the 2014 Off the Beaten Track workshop. This work won first place at the 2014 PLDI Student Research Competition, a best paper award at OOPSLA 2014, and a 2015 Outstanding Synthesis Award at the University of Massachusetts Amherst. UVM PhD student Michael McConnell resurrected this work with an interest in applying quantitative information flow to surveys to protect potentially vulnerable populations.

Current status

As a result of serendipitous opportunitites afforded by my UVM colleagues Joe Near, I had returned to this work due to the confluence of student interest in privacy, natural language processing, and biased data collection. Upon leaving UVM, I had to put this work on hold, but am keen to work on it again.

Future projects

The release of tools that leverage large langauge models have made some of the more labor-intenstive aspects of this work potentially much lighter. If you would like to discuss further, please reach out.