Week 8& 9: Reasoning and Problem Solving

Litt, A., Eliasmith, C., & Thagard, P. (forthcoming). Neural affective decision theory: Choices, brains, and emotions. Cognitive Systems Research.

Evans, J.S.T.B. (2003). In two minds: Dual process accounts of reasoning. Trends in Cognitive Science, 7, 454-459.

Thagard, P. (2007). Abductive inference: From philosophical analysis to neural mechanisms. In A. Feeney & E. Heit (Eds.), Inductive reasoning: Experimental, developmental, and computational approaches (pp. 226-247). Cambridge: Cambridge University Press.


This week’s readings focus on the process of reasoning and decision making. Throughout each article and chapter, a common theme is reported: both reasoning and decision making appear to be the result of a final solution reached through the interaction of dual processing streams, one involving emotional processing and the other involving more cognitive processing. Evans (2003) presents a dual processing theory of reasoning, whereby two systems essentially “compete” for the final solution. “System 1” processing represents rapid, automatic processing, representing concepts and beliefs that are formed through associative learning. It is through this system that innate and instinctual behaviors are accessed. Evans proposes the end result of the rapid and automatic processing by this system is what becomes available to consciousness. System 2, by contrast, represents much slower, methodical processing, making use of working memory systems to elaborate upon information from System 1, engaging in more sophisticated hypothetical thinking and forecasting, constructing mental models and analyzing possible outcomes. Through this process, System 2 essentially has the capacity to override System 1. Evans presents examples from studies in which syllogisms are used to evaluate the relationship between beliefs and analytical deductive reasoning. In one type of study, participants are asked to endorse only those conclusions that logically follow a preceding premise regardless of their beliefs. Results show participants have a very difficult time overriding prior beliefs, and show belief bias in their endorsement of conclusions, rejecting otherwise logically deduced solutions. For example, in the syllogism “No nutritional things are inexpensive; Some vitamin tablets are inexpensive; Therefore, some vitamin tablets are not nutritional,” participants demonstrated difficulty endorsing this conclusion regardless of the fact it rationally follows the previous premises, having a difficult time “buying” that we can conclude some vitamins are not nutritional based on the fact they are inexpensive. If instructions to participants emphasize the importance of endorsing conclusions only on the basis of their logical merit, participants are able to do it, but only with effort. Evans proposes these studies are representing the process of System 2 inhibition and override of System 1. If System 2 requires working memory and other higher cognitive processes in order to override System 1, then measures of intelligence ought to correlate with ability to inhibit System 1 by System 2. This has indeed been demonstrated, with higher IQ scores correlating with greater ability to find correct solutions in reasoning tasks. In short, the greater cognitive capacity an individual has, the better able they are to go beyond “gut reactions” to problems and find other possible solutions. This is analogous to the stereotype of the reckless, emotional decision-maker versus the cool, calm, collected and calculated one – Inspector Clouseau versus James Bond if you will.

In Thagard’s article, he provides arguments for a neural account of abductive reasoning. Abductive reasoning refers to inference involving the generation and evaluation of explanatory hypotheses. Thagard argues that the process of abductive reasoning is inherently emotional, based on the fact that reasoning occurs first when something is puzzling, which is resolved when a target explanatory solution is arrived at. Both the puzzling nature and the satisfaction with the explanatory solution are in essence emotional events. He suggests that the ability to find causal relations begins with very early perceptual processing, as demonstrated by studies showing infants as young as 2.5 months expect that a stationary object will be displaced when hit by a moving object. Thagard proposes there is a neurally-encoded image schema that establishes the causal relationship tying the neural structure representing hypothesis with the neural structure representing the target explanation. Abductive inference is the “transformation of representational neural structures that produces neural structures that provide causal explanations.” Abductive inference does not only include verbal-linguistic processing but also inference from multiple perceptual modalities (such as deducing from seeing a scratch on your car in a supermarket parking lot and a shopping cart nearby that the shopping cart caused the scratch). All types of inference are inherently emotional in that what motivates one to find a causal explanation is the emotional thrust of puzzlement, and what represents a solution is the satisfaction that solution elicits. Here again, we see the interaction between emotion and cognition.

Litt, Eliasmith, and Thagard provide an interesting account of the role of emotion in decision making. Decision-making involves the weighting of various response choices and their potential consequences. As discussed earlier in Week 6, this involves both emotional and contextual information, implicating VMPFC, amygdala, and hippocampus. The current article extends upon this and demonstrates through neurocomputational modeling how amygdala activation (representing emotional salience) influences ongoing response selection. In essence, the greater emotional arousal generated by stimuli, the greater the subjective value placed on the stimuli by the OFC. Valuations are exponentially dampened or intensified depending upon the lowered or heightened state of arousal. The authors provide equations representing this process, demonstrating how the level of amygdala activation can in essence cancel out OFC responses. Greater negative predictions elicit higher levels of arousal, and there is greater aversion to potential losses than gains in predicted outcomes. The authors go on to present fascinating accounts for the way in which framing a problem can influence decision making. Potential for loss is more arousing than potential for gain. Therefore, the way a problem is presented, emphasizing overall losses as opposed to overall gains, influences which decision is made. For example, studies by Tversky & Kahneman (1981, 1986) found when given a choice of two plans to control an outbreak expected to kill 600 people, participants were inclined to choose a plan that would result in 200 people being saved but reject a plan resulting in 400 people being killed. Objectively, both of these choices are exactly the same (200 people live, 400 people died), but when presented as an opportunity to save people the choice was more desirable than when presented with the opportunity to kill people. The same framing phenomenon occurs in the famous trolley-footbridge dilemma (Greene et al., 2002, 2004), wherein participants are more likely to chose to push a button releasing a runaway train car carrying multiple people, risking multiple peoples’ lives, than choosing to push one person in front of the runaway train, killing that person but saving the rest. Even though more people will likely die in the first option, the distance between the action of pushing a button and that action causing death is greater than the distance between making physical contact with an individual and causing death. The latter elicits far greater amygdala and OFC activation than the former, suggesting greater emotional salience. Another aspect of framing explains why people sometimes make choices that are objectively less valuable but hedonically more valuable. The authors give the example that winning $20 feels like a gain when the comparison is winning only $1, whereas winning $20 feels like a loss when the comparison is winning $100. It is objectively the same outcome, $20 is $20, but one outcome is more desirable than the other. The authors suggest the difference in desirability is the result of the distance between the actual outcome and the expected outcome. If you expect to earn $100, $20 feels like a loss. However, if you expect to win $1, $20 is a gain.
The article by Litt et al. maps well onto the article by Evans, wherein we can assume the hedonic value and emotional contribution to a decision is a result of “System 1” processes, resulting from prior learned associations and innate beliefs (such as killing another human being is bad). The degree to which the emotional aspects of a decision or reasoning process win out is related to the degree to which further elaboration and hypothesizing about a possible solution generated through “System 2” processes override System 1 contributions. Regardless, it appears we cannot “escape” bottom-up, affect-driven influences on what would otherwise be construed as a cognitive process.