A. David Redish > Quotes

“careful scientific studies have revealed that a lot of conscious “decisions” that we think we make are actually rationalizations after the fact.”
― A. David Redish, The Mind within the Brain: How We Make Decisions and How those Decisions Go Wrong

“In the Redish model, the excess dopamine provides additional value, no matter what. Marks et al. (2010) directly tested this hypothesis in an elegant experiment, where rats were trained to press two levers for a certain dose of cocaine (both levers being equal). One lever was then removed and the other provided smaller doses of cocaine. The Redish theory predicts that the second lever should gain value, while expectation of homeostatis theories would predict that the second lever should lose value (because animals would learn the second lever was providing smaller doses). The Marks et al. data was not consistent with the Redish excess-delta model. However, a key factor in drug addiction is that not everyone who takes drugs loses control over their drug use and becomes an addict. Studies of drug use in both human and nonhuman animals suggest that most animals in self-administration experiments continue to show elasticity in drug-taking, stopping in response to high cost, but that a small proportion (interestingly similar to the proportion of humans who become addicted to drugs) become inelastic to drug-taking, being willing to pay excessive costs for their drugs. One possibility is that the homeostatic models are a good description of nonaddicted animals, which have a goal of maintaining a satiety level, but that addiction is different.”
― A. David Redish

“Multi-system models suggest that addiction is a question of harmful dysfunction - dysfunction (vulnerabilities leading to active failure modes) within a system that causes sufficient harm to suggest we need to treat it. They permit both behavioral and pharmacological drivers of addiction.
The suggestion that different decision-making systems can drive behavior provides a very interesting treatment possibility, which is that one could potentially use one decision-system to correct for errors in another. Three computational analyses of this have been done - changing discounting rates with episodic future thinking, analyses of contingency management, and analyses of precommitment.”
― A. David Redish

“The Redish model also predicted that drugs would not show Kamin blocking. Kamin blocking is a phenomenon where animals don't learn that a second cue predicts reward if a first clue already predicts it. This phenomenon is well-described by value-prediction error (VPE) - once the animal learns that the first cue predicts the reward, there is no more VPE (because it's predicted!) and the animal does not learn about the second cue. Redish noted that because drugs provided dopamine, and dopamine was hypothesized to be that VPE delta signal, that when drugs were the 'reward', there was always VPE. Thus, drug outcomes should not show Kamin blocking. The first tests of this did not conform to the prediction - animals showed Kamin blocking, even with drug outcomes. However, Jaffe et al. (2014) wondered whether this was related to the subset problem - that only some animals were actually overvaluing the drug. Jaffe et al. tested rats in Kamin blocking for food and nicotine. All rats showed normal Kamin blocking for food. Most rats showed normal Kamin blocking for nicotine. But the subset of rats that were high responders to nicotine did not show Kamin blocking to nicotine, even though they did to food, exactly as predicted by the Redish model.”
― A. David Redish

“Building on the anatomical data known to drive the typical shift from planning to procedural decision systems, Piray et al. (2010) proposed a computational model in which drugs disrupted the planning-valuation systems and accelerated learning in the procedural-valuation systems. This model suggested that known changes in dopaminergic function in the nucleus accumbens as a consequence of chronic drug use could lead to overly fast learning of habit behaviors in the dorsal striatum and would provide a shift from planning to habit systems due to changes in valuation between the two systems.”
― A. David Redish

“One of the classical descriptions of addictions is based on the observation that addicts will continue to use even in the face of high costs. This can be quantified through the economic concept of elasticity as a measure of how much one's willingness to buy something changes by its cost. Things that diminish slowly by cost are inelastic. Researchers have suggested that drugs are fundamentally inelastic: as costs increase, the number of rewards paid for decrease less than they should. Of course, there are many things that are inelastic that are not considered addictive - oxygen, for example (where the withdrawal symptoms are particularly traumatic), but also some behaviors continued even in the face of high costs are celebrated, such as Kerri Strug's 1996 Olympic vault performed on a sprained ankle, or Osip Mandelstam continuing to write poetry even after Stalin had thrown him in the gulag for it.”
― A. David Redish

“The fact that addicts show fast discounting functions with preferences that change over time suggests two interesting related treatments: bundling and precommitment.
Bundling is a process whereby multiple rewards are grouped together so as to calculate the value of the full set rather than each individually.
A similar process is that of precommitment, where a subject who knows in advance that if given a later option, the subject will take the poor choice, prevents the opportunity in the first place. Economically, precommitment depends on the hyperbolic discounting factors that lead to preference reversals. Although many experiments have found that the average subject shows hyperbolic discounting, individuals can show large deviations from good hyperbolic fits. Computationally, an individual's willingness to precommit should depend on the specific shape of their discounting function.”
― A. David Redish