The goal of our work is to investigate best practices for integrating machine learning into human decision making. There are many different ways to incorporate machine learning in the decision making process. The above image illustrates a spectrum between full human agency, where humans make decisions entirely on their own, and full automation, where machines make decisions without human intervention. The amount of information from machine learning models gradually increases along the spectrum. Note that assistance is in the form of offering explanations and predictions. For more details on our work, please check out our papers here.

This website allows you to do following tasks with different levels of machine assistance along the spectrum:

• Deception: detect deceptive hotel reviews;
• Bios: guess someone's occupation from a short biography;
• Recidivsm I: judge whether a defendant will recidivate;
• Recidivsm II: judge whether a defendant will violate terms of pretrial release;

By using this website, you agree to share your responses for research purposes.

Click on any button to select a task.