I study how people interact with conversational robots in civic settings like public libraries and classrooms. My work explores how physical embodiment shapes perceived judgment and engagement, investigating where robots meaningfully expand accessibility and where human relationships should never be replaced.
Sean Collins, N. C. Morris, M. Diamond, M. Wills, Jevon Lipsey, Tom Williams
Homelessness is a complex, prevalent, and socially entangled issue that affects a multitude of aspects within an individual's life. While numerous organizations work together to provide necessary services and support to unhoused people, many of the needs of unhoused people remain unmet. For roboticists seeking to work with this population, this raises two key research questions: what needs remain unmet in this community, and which of those needs might be addressable through robotic solutions? To answer these research questions, we conducted a five phase research process. First, we conducted interviews with those who work and volunteer with communities of unhoused people through local organizations. Second, we analyzed the transcripts of those interviews through the lens of the Existence, Relatedness, and Growth theory of needs. Third we conducted a prevalence analysis to prioritize these needs from the perspective of service providers. Fourth, we conducted a design fiction workshop to envision possible robotic solutions to unmet needs and to prioritize the needs undergirding those solutions from the perspective of designers. Finally, we analyzed the three remaining needs identified through this process through the lens of prior interviews with unhoused community members. Our results reveal a complex network of interlocking unmet needs and suggest that social robots might best aid unhoused people through the nonjudgmental provision of information about available resources and social services.
Knowledge bases traditionally require manual optimization to ensure reasonable performance when answering queries. We build on previous neurosymbolic approaches by improving the training of an embedding model for logical statements that maximizes similarity between unifying atoms and minimizes similarity of non-unifying atoms. In particular, we evaluate different approaches to training this model.
Y. Zhang, Y. White, D. Clark, J. Sanchez, Jevon Lipsey, A. Hirst, J. Heflin
Neural networks can be trained to rank the choices made by logical reasoners, resulting in more efficient searches for answers. A key step in this process is creating useful embeddings, i.e., numeric representations of logical statements. This paper introduces and evaluates several approaches to creating embeddings that result in better downstream results. We train embeddings using triplet loss, which requires examples consisting of an anchor, a positive example, and a negative example. We introduce three ideas: generating anchors that are more likely to have repeated terms, generating positive and negative examples in a way that ensures a good balance between easy, medium, and hard examples, and periodically emphasizing the hardest examples during training. We conduct several experiments to evaluate this approach, including a comparison of different embeddings across different knowledge bases, in an attempt to identify what characteristics make an embedding well-suited to a particular reasoning task.
