My research focuses on something I call adversarial interaction.
Adversarial interaction is when people use online platforms strategically and in unexpected ways. But why would a user want to interact adversarially?
Users and platforms are often in conflict. What a user wants may not be good for business. A product’s interface is a choice architecture where some outcomes are systematically favored over other theoretically-achievable outcomes. Entering into someone else’s choice architecture has ramifications for user autonomy—will they be able to make the choices they want to make when interacting with the system or will there be engineered friction to realizing certain outcomes?
Our core insight is that the design of most software products is contestable—a user downloads code and data from a platform and usually executes the code as instructed on the data, as the platform expects. However, this code, and thus the product design and the underlying choice architecture, is often just a suggestion. A user executing code in unexpected ways might co-author the product design in order to achieve outcomes they prefer more often.
We look at user-platform conflicts using tools from the fields human-computer interaction and security, incentives in online systems using tools from game theory and mechanism design, and ways that users can use software to modify their interactions to favor themselves rather than platforms in these conflicts.
The most obvious example of adversarial interaction is adblocking. Another example would be a user who wants to spend less time on Facebook, in spite of how bad this decision might be for the company’s business model. Facebook is designed to be addictive, but rendering the site based on their design is a suggestion, not actually required. A user might use a tool where they specify how much they’d like to use Facebook. Once this usage threshold is reached, the tool might degrade addictive design patterns in Facebook to make the user experience more frustrating—by washing out colors, blurring text, and making the site load more slowly.
Overall, we want to know what can be achieved by viewing the interface as a contested space, by what means and at what cost?