Uploaded February 2023 | Updated September 2026, 3 days ago
Abstract: “Ask not what's inside your head, but what's your head inside of” — W.M. Mace, 1977
In contrast to classical cognitive science which studied brains in isolation, ecological cognitive science accounts focus on the role of the body and external environment in shaping cognition. In this talk I will
describe how ecological approaches informed my PhD thesis, as well as the new research directions I’m looking to pursue. The first part will focus on ecological approaches to grounded natural language understanding (NLU) research. Where mainstream research often focuses on designing new models and optimization methods while treating learning environments as given, we explored the potential of environment design for improving data collection and model development. We developed a novel training and annotation approach for procedural text, using text-based games (TBG) environments to ensure annotation validity and automate tedious state-tracking. We also used TBGs to construct a new configurable question answering benchmark for measuring the progress of large language models on challenging commonsense reasoning tasks. We then leveraged the richer supervision provided by TBG environments to develop Breakpoint Transformers, a novel approach to modeling intermediate semantic information in long narrative or procedural texts. The second part will focus on ecological (human) collective intelligence (CI) theory, and how it can inform the design of AI-augmented “social sensemaking environments” for knowledge workers like scientists. Finally, I’ll discuss how these directions could be integrated with Semantic Scholar to help researchers navigate information overload and form new collaborations.
Bio: Ronen Tamari is a final-year PhD student at the Hebrew University of Jerusalem. His work broadly focuses on the intersection of grounded NLP and contemporary cognitive science theories (embodied, extended, ecological). He is particularly interested in the role of learning environments, and how their structure and affordances shape artificial, human and collective agents’ behavior. Ronen is also interested in building technology to empower communities: he previously led the collective intelligence (CI) research team at DAOstack and co-founded Common SenseMakers, an R&D collective for CI technology.
Abstract: “Ask not what's inside your head, but what's your head inside of” — W.M. Mace, 1977
In contrast to classical cognitive science which studied brains in isolation, ecological cognitive science accounts focus on the role of the body and external environment in shaping cognition. In this talk I will
describe how ecological approaches informed my PhD thesis, as well as the new research directions I’m looking to pursue. The first part will focus on ecological approaches to grounded natural language understanding (NLU) research. Where mainstream research often focuses on designing new models and optimization methods while treating learning environments as given, we explored the potential of environment design for improving data collection and model development. We developed a novel training and annotation approach for procedural text, using text-based games (TBG) environments to ensure annotation validity and automate tedious state-tracking. We also used TBGs to construct a new configurable question answering benchmark for measuring the progress of large language models on challenging commonsense reasoning tasks. We then leveraged the richer supervision provided by TBG environments to develop Breakpoint Transformers, a novel approach to modeling intermediate semantic information in long narrative or procedural texts. The second part will focus on ecological (human) collective intelligence (CI) theory, and how it can inform the design of AI-augmented “social sensemaking environments” for knowledge workers like scientists. Finally, I’ll discuss how these directions could be integrated with Semantic Scholar to help researchers navigate information overload and form new collaborations.
Bio: Ronen Tamari is a final-year PhD student at the Hebrew University of Jerusalem. His work broadly focuses on the intersection of grounded NLP and contemporary cognitive science theories (embodied, extended, ecological). He is particularly interested in the role of learning environments, and how their structure and affordances shape artificial, human and collective agents’ behavior. Ronen is also interested in building technology to empower communities: he previously led the collective intelligence (CI) research team at DAOstack and co-founded Common SenseMakers, an R&D collective for CI technology.
