Ecology
Agents operate in an environment but most experiments construct sterile, artificial environments to exert epistemic control. So is looking at the environment worthwhile or can we understand cognition by looking at agents in fixed environments?
Primary Readings
Everyone should read these and be prepared to discuss.
| Anderson, J. R., & Schooler, L. J. (1991) | Reflections of the environment in memory. Psychological Science, 2(6), 396-408. |
| Shepard, R. N. (1987) | Toward a universal law of generalization for psychological science. Science, 237(4820), 1317-1323. |
Secondary Readings
The presenter should read and incorporate at least two of these:
| Griffiths, T. L., & Tenenbaum, J. B. (2006). | Optimal predictions in everyday cognition. Psychological Science, 17(9), 767-773.This research article explores the optimality of human cognition in everyday situations and whether cognitive judgments align with optimal statistical principles. The study investigates various domains, including life spans, movie grosses, poem lengths, baking times, waiting times, and reigns of pharaohs. The results demonstrate that people’s predictions in most domains closely match the optimal Bayesian predictions based on empirical prior distributions. This one could easily be in the Rationality" topic, but I put it here to highlight that effective inductive biases are effective because they reflect actual environmental statistics. |
| Feldman, J. (2012) | Symbolic representation of probabilistic worlds. Cognition, 123(1), 61-83.The article discusses the use of symbolic representation in cognitive science, specifically in probabilistic worlds. The author introduces the concept of modality and argues that symbolic representation depends on the probabilistic structure of the environment. The paper proposes that modal environments, which consist of mixtures of narrow component distributions, can be represented symbolically with minimal information loss. The author also explores the relationship between discrete symbols and continuous variables, as well as the challenges of working with mixtures. The article discusses the implications of modality in the environment and how it affects an observer’s ability to represent and comprehend it. It concludes that environments with multiple modes or peaks can be effectively represented by symbols, while arbitrary environments cannot. The paper also examines the reduction of dimensions and how modal environments can be approximated by symbolic variables. Overall, the paper aims to show that symbolic representations in cognitive science are effective in capturing the probabilistic structure of modal worlds. |
| Harnad, S. (1990) | The symbol grounding problem. Physica D: Nonlinear Phenomena, 42(1-3), 335-346.This paper by Stevan Harnad, published in 1990, explores the symbol grounding problem and the role of connectionism in cognitive modeling. It suggests that symbols can be grounded in nonsymbolic representations, such as iconic and categorical representations, in order to give them meaning. The paper discusses the strengths and weaknesses of both connectionism and symbolism, and proposes a hybrid model that combines these approaches to address the limitations of each. It also discusses the transition from behaviorism to cognitivism in the modeling of the mind, and explores the defining characteristics of a symbol system. The paper concludes that addressing the symbol grounding problem is crucial for understanding complex cognitive capacities, and that a cooperative approach between connectionism and symbolism is necessary for a comprehensive cognitive model. Overall, the paper highlights the challenges and potential solutions in giving meaning to symbols within a formal symbol system, and discusses the importance of grounding symbols in nonsymbolic representations. |
Questions under discussion
- Is the environment worth studying?
- What is the target for the grounding problem?
