Taming Uncertainty by Ralph Hertwig (See posts Dialectical Bootstrapping and Harnessing the Inner Crowd.), Timothy J Pleskac (See post Risk Reward Heuristic.), Thorsten Pachur (See post Emotion and Risky Choice.) and the Center for Adaptive Rationality, MIT Press, 2019, is a new compendium that I found accidentally in a public library. There is plenty of interesting reading in the book. It takes the adaptive toolbox approach as opposed to the Swiss Army Knife . The book gets back cover raves from Cass Sunstein (See posts Going to Extremes, Confidence, Part 1.), Nick Chater, and Gerd Gigerenzer (See post Gigerenzer–Risk Saavy, and others.). I like the pieces, but not the whole.
Hertwig et. al. start with Simon’s bounded rationality as a psychologically realistic model as opposed to the more Olympian models that the authors see as ignoring our limited cognitive resources. They build on bounded rationality first with heuristics– simple adaptive tools that with modest effort and computation can give good performance. To heuristics and the adaptive toolbox, they have added learning through experience, social intelligence, and aggregation of information. They see this repertoire as helping people navigate uncertainty. They are definitely part of the Gigerenzer heuristics school. Gigerenzer’s favorite and mine is the gaze heuristic (See post Bounded Rationality-the Adaptive Toolbox.). They differentiate their ideas from Kahneman (See post Cumulative Prospect Theory) and Tversky who looked at heuristics and biases often noting how humans fell short of rational choice models. Hertwig (p359-360) states:
By contrast, our view is that heuristics… offer a different way to deal with uncertainty that people…commonly encounter in their daily lives….In this view, the essence of rational behavior consists in how an organism can adapt in order to achieve its goals under the constraints posed by both the environment and its own cognitive limitations.
Hertwig et al define rationality as Hammond’s correspondence as opposed to his coherence (See post Beyond Rationality – Part 1.). This is based on the concept of ecological rationality whereby the environment and agent are not divorced from one another. Thus, domain-specific versus domain-general methods are required, and according to the authors, this suggests that there cannot be a single universal and domain general tool. Hertwig et al do not “see cognition in an uncertain world in terms of a single computationally powerful and optimizing prediction machine,” and they specifically mention Andy Clark who definitely does believe.
In response to these ideas, I must note that Kenneth Hammond did not go all in with correspondence. He saw both correspondence and coherence as necessary. Second, Hertwig has not convinced me that Clark’s prediction machine cannot coexist with his heuristics. Another reason is mentioned by Hertwig:
Our results suggest an interesting twist to a key question: How does the mind select a heuristic from the adaptive toolbox? The simulations of choice heuristics…demonstrate one thing: there is great variability in the performance of heuristics and choosing the wrong one in a given environment may prove costly. How does the mind figure out which heuristic to select? ….There is, however another answer. The good average performance of the equiprobable heuristic under conditions of highly limited knowledge…suggests that some heuristics may offer good fallback options when an informed strategy selection is not possible. (Taming Uncertainty, p 47-48.)
This seems ridiculous to me. It is almost saying that the equiprobable heuristic is the backup universal and domain general tool which they have earlier said cannot exist. Maybe they would like Clark’s minimizing prediction error better if he called it a heuristic. Clark admits to having been more or less an adaptive toolbox guy In the blog “The Brains” on December 14, 15, 16, and 17, 2015:
Is the human brain just a rag-bag of different tricks and stratagems, slowly accumulated over evolutionary time? For many years, I thought the answer to this question was most probably ‘yes’. Sure, brains were fantastic organs for adaptive success. But the idea that there might be just a few core principles whose operation lay at the heart of much neural processing was not one that had made it on to my personal hit-list.
The idea has now made Clark’s list. In the chapter: “Embodied Prediction” in T. Metzinger and J.M. Windt (Eds) Open MIND: 7(T) Frankfurt am Main: MINDGroup, 2015, he easily incorporates the gaze heuristic into his organizing principles. According to Clark, altering the distribution of precision weightings amounts to altering the “simplest circuit diagram” for current processing. This suggests a new angle upon the outfielder’s problem. Already-active neural predictions and simple, rapidly-processed perceptual cues must work together to determine a pattern of precision weightings for different prediction-error signals. This creates a pattern of effective connectivity (a temporary distributed circuit) and, within that circuit, it sets the balance between top down and bottom-up modes of influence. In the case at hand, however, efficiency demands selecting a circuit in which visual sensing is used to cancel the optical acceleration of the fly ball. This means giving high weighting to the prediction errors associated with cancelling the vertical acceleration of the ball’s optical projection, and not caring very much about anything else. Apt precision weightings here function to select what to predict at any given moment.
Clark states that this mechanism (the so-called ‘precision-weighting’ of prediction error) provides a flexible means to vary the balance between top-down prediction and incoming sensory evidence, at every level of processing. Implemented by multiple means in the brain (such as slow dopaminergic modulation, and faster time-locked synchronies between neuronal populations), flexible precision-weighting makes these architectures fluid and content responsive (See posts Embodied(Grounded) Prediction(Cognition) and the Mixed Instrumental Controller).