Causal Networks with Selectively Influenced Components
Abstract
In work supervised by Dr. Schweickert, a method for inferring a processing tree to account for response probability was generalized to allow for multiple response classes, parameters such as rates not bounded above by 1, and experimental factors having effects at more than one vertex. A further generalization allows a processing tree to be inferred from joint analysis of reaction time and accuracy. Processing trees were inferred for immediate serial recall data and data on sleep and interference. In the part under Dr. Dzhafarov's supervision, a general mathematical theory of selective influences was elaborated (which input influences which of probabilistically interdependent random outputs); the Joint Distribution Criterion was formulated in complete generality; a theory of pseudo-quasi-metrics was constructed to be used to test for selectiveness of influences; a Linear Feasibility Test for selective influences with finite-valued random outputs was constructed; a formal equivalence of selective influences with the issue of quantum entanglement in physics was established, with non-commuting measurements in quantum physics paralleling the mutually exclusive values of inputs (external factors) in behavioral sciences.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 29, 2012
- Accession Number
- ADA565335
Entities
People
- E. N. Dzhafarov
- Janne V. Kujala
- R. Schweickert
Organizations
- Purdue University