Elizabeth Bondi, Harvard University: “Imagery and Strategic Reasoning: Making Decisions in Conservation and Sustainability with Imperfect Data”

Position: PhD Candidate

Current Institution:  Harvard University

Abstract: Imagery and Strategic Reasoning: Making Decisions in Conservation and Sustainability with Imperfect Data

My research interests lie in the area of artificial intelligence (AI) for social impact, specifically at the intersection of multi-agent systems and computer vision, to support decision-making in important real-world, resource-constrained efforts in sustainability and public health. AI for social impact considers end-to-end systems ranging from data to deployment. The key insights of my work are that it is crucial to (1) incorporate uncertainty at all stages of this pipeline, rather than only ending it at the initial data sensing stage as in most previous work, and (2) use different techniques specialized for each stage. For example, to prevent illegal wildlife poaching, previous research has independently focused on addressing uncertainty only in the object detection phase in conservation drone imagery, but has not considered uncertainty in downstream defender-attacker security games with signaling. Unfortunately, ignoring uncertainty as in prior work in security games with signaling leads to large losses for the defender. In contrast, my work addresses uncertainty beginning in the sensor stage, through to the higher-level reasoning stage of defender-attacker security games with signaling. We introduce novel techniques, such as introducing additional randomized signalling in the security game, to handle uncertainty appropriately at this stage in the pipeline, thereby reducing losses to the defender. We show similar reasoning is important in public health, where we would like to predict disease prevalence with few ground truth samples in order to better inform policy, such as optimizing resource allocation. In addition to modeling such real-world challenges holistically, we must also work with all stakeholders in this work, including by making our field more inclusive through efforts like my nonprofit startup, Try AI.


Elizabeth Bondi is a PhD candidate in Computer Science at Harvard University advised by Prof. Milind Tambe. Her research interests include multi-agent systems remote sensing computer vision and deep learning especially applied to conservation and sustainability. Among her awards are Best Paper Runner up at AAAI 2021 Best Application Demo Award at AAMAS 2019 Best Paper Award at SPIE DCS 2016 and an Honorable Mention for the NSF Graduate Research Fellowship Program in 2017.