Insight and strategy in multiple-cue learning.

Abstract

In multiple-cue learning (also known as probabilistic category learning) people acquire information about cue-outcome relations and combine these into predictions or judgments. Previous researchers claimed that people can achieve high levels of performance without explicit knowledge of the task structure or insight into their own judgment policies. It has also been argued that people use a variety of suboptimal strategies to solve such tasks. In three experiments the authors reexamined these conclusions by introducing novel measures of task knowledge and self-insight and using "rolling regression" methods to analyze individual learning. Participants successfully learned a four-cue probabilistic environment and showed accurate knowledge of both the task structure and their own judgment processes. Learning analyses suggested that the apparent use of suboptimal strategies emerges from the incremental tracking of statistical contingencies in the environment.Keywords: multiple-cue learning, self-insight, strategy, rolling regression, implicit versus explicit learning A fundamental goal of cognition is to predict the future on the basis of past experience. In an uncertain world this involves learning about the probabilistic relations that hold between the available information and an outcome of interest and integrating this information into a singular judgment. Thus, a stockbroker draws on various market indicators to predict the future price of a share, and a race-course pundit uses factors such as form and track condition to pick the likely winner of a race.The underlying structure of this kind of prediction is captured in the multiple-cue learning framework (Brunswik, 1943;Hammond, 1955), which focuses on how people learn from repeated exposure to probabilistic information. This paradigm, also known as probabilistic category learning, has been applied in numerous areas of psychology, including human judgment (Brehmer, 1979;

Fear relevancy, strategy use, and probabilistic learning of cue-outcome associations

“…Learning strategies were evaluated using mathematical models to fit each participant's data to the ideal data if a subject were reliably following a particular strategy using procedures detailed by Lagnado et al (2006). Separate analyses were conducted for each run in each group to assess changes in strategy use across the experiment.…”

“…Separate analyses were conducted for each run in each group to assess changes in strategy use across the experiment. The performance of individual participants was compared with that of an ideal participant performing one of three different strategies: (1) "simple strategies" encompassing both singleton and one-cue strategies, (2) "complex strategies" including both multimatch and multimax strategies, or (3) "no identifiable strategy" (for details, see Lagnado et al 2006). A leastmeans-squared estimate was computed to evaluate the likelihood that each participant's pattern of responses followed a certain strategy across each 50-trial run.…”

“…Therefore, strategy use (simple, complex, nonidentifiable) was mathematically modeled according to the procedures described in Lagnado et al (2006) (see Materials and Methods), and the proportion of participants using each strategy was characterized for each group and run. For statistical purposes, subjects using multimax and multimatch strategies were collapsed into a "complex strategy" category, and subjects using singleton and one-cue strategies were collapsed into a "simple strategy" category.…”

“…To accomplish this objective, each participant's data is mathematically modeled across trial blocks to determine the goodness-of-fit to an "ideal" responder following particular response patterns. Least-mean-square estimates indicate that participants use at least three classes of strategies varying in optimality: (1) no identifiable strategy, (2) simple strategies involving the use of one cue to make predictions, and (3) complex strategies involving the use of multiple cues and knowledge of the underlying probabilistic structure (Gluck et al 2002;Lagnado et al 2006;Meeter et al 2006).Probabilistic learning of this sort is likely to be important for assessing the motivational and emotional relevance of stimulus contingencies in real-life scenarios, which are often nondeterministic. However, it is not known how behavioral performance or strategy use on PCL tasks is impacted by varying the salience of the outcomes.…”

Models of probabilistic category learning in Parkinson’s disease: Strategy use and the effects of L-dopa

“…Both healthy (Lagnado et al, 2006) and amnesic participants have shown explicit insight into the cue-outcome contingencies, thus indicating access to representations of the task environment which should be absent in purely habitual learning. Indeed, there is now considerable evidence that PCL relies at least partially (Meeter et al, 2006), if not wholly (Lagnado et al, 2006;Newell et al, 2007;Price, 2009), on explicit processes. Furthermore, a recent study failed to replicate the results of Shohamy et al (2004a).…”