Schedule

It is generally believed that fairness and privacy awareness are conflicting goals in algorithmic design. We, however, show that it is possible to jointly consider the two in some optimization settings. The corresponding results are then discussed in the context of robotic decision-making algorithms, which are increasingly facing legal, ethical and social constraints as robots interact more with human peers. I look into how to navigate robots in human work environments while respecting possible fairness and privacy requirements. Such scenarios are increasingly common, for instance, when robots transport sensitive objects  in crowded spaces. To address this, we propose a new framework for mobile robot navigation that leverages vision language models in adaptive path planning. Experimental results demonstrate that our framework is effective in human-robot shared public environments.

Ranking algorithms have recently come under scrutiny for preventing minority groups from reaching higher ranking slots in applications like search and recommendation, thus reducing their visibility. In this talk, I will describe our recent work in diagnosing when and how algorithms that use network information may further bias against minority groups. We focus on two famous algorithms, PageRank and HITS, and analyze them empirically and theoretically, using a generative network model with multiple communities. We find that HITS amplifies pre-existing bias in homophilic networks, as compared to PageRank. We find the root cause of bias amplification in HITS to be the level of homophily present in the network through a novel theoretical analysis involving network models that are predisposed to bias against minority groups. This work is joint with Augustin Chaintreau and Nelly Litvak and was published at The Web Conference ’24.

Two-sided matching markets have long existed to pair agents in the absence of regulated exchanges. A common example is school choice, where a matching mechanism uses student and school preferences to assign students to schools. In such settings, forming preferences is both difficult and critical. Data-driven decision support systems can predict future outcomes based on historical data to help students form their preferences. Although often deployed together, these matching and prediction mechanisms are usually analyzed separately. But what happens if we consider the market as a whole? In this talk, I present a new type of strategic behavior of schools targeting the prediction mechanism by leveraging the repeated nature of the market. This strategic behavior is not merely an attack on the reported data but a change in the very school-student interaction. I will discuss when such strategies are optimal and what their consequences are. In the context of our workshop, this is an example of how introducing and improving data-driven decision support systems can incentivize new strategic behaviors that ultimately cause social inequalities independent of individual potential.

We study fairness in social influence maximization, whereby one seeks to select seeds that spread a given information throughout a network, ensuring balanced outreach among different communities (e.g. demographic groups). In the literature, fairness is often quantified in terms of the expected outreach within individual communities. In this talk, we demonstrate that such fairness metrics can be misleading since they overlook the stochastic nature of information diffusion processes. When information diffusion occurs in a probabilistic manner, multiple outreach scenarios can occur. As such, outcomes such as “In 50% of the cases, no one in group 1 gets the information, while everyone in group 2 does, and in the other 50%, it is the opposite”, which always results in largely unfair outcomes, are classified as fair by a variety of fairness metrics in the literature.We tackle this problem by designing a new fairness metric, mutual fairness, that captures variability in outreach through optimal transport theory. We propose a new seed-selection algorithm that optimizes both outreach and mutual fairness, and we show its efficacy on several real datasets. We find that our algorithm increases fairness with only a minor decrease (and at times, even an increase) in efficiency.