How to Make Decisions:
Coherence, Emotion, and Practical Inference
Paul Thagard
Philosophy Department
University of Waterloo

Thagard, P. (2001). How to make decisions: Coherence, emotion, and practical inference. In E. Millgram (Ed.), Varieties of practical inference . Cambridge, MA: MIT Press. 355-371.

Students face many important decisions: What college or university should I attend? What should I study? What kind of job should I try to get? Which people should I hang out with? Should I continue or break off a relationship? Should I get married? Should I have a baby? What kind of medical treatment should I use? A theory of practical reasoning should have something to say about how students and other people can improve their decision making.

I regularly teach a first-year course on critical thinking intended to help students improve their reasoning about what to believe and about what to do. After spending about two thirds of the course on ways of improving judgments about the truth and falsity of controversial claims in areas such as medicine and pseudoscience, I devote the last third to practical reasoning, with the focus on how people can make better decisions. I discuss both the kinds of erroneous reasoning that decision makers commonly fall into, and some systematic models that have been proposed by psychologists, economists, and philosophers to specify how people should make decisions.

Many students in the course dislike these models, and resist the claim that using them is preferable to making decisions simply by intuition. They trust their "gut feelings" more than they trust the analytical methods that require a systematic and mathematical comparative assessment of competing actions that satisfy multiple criteria. The textbooks I use (most recently Gilovich 1991, Russo and Schoemaker 1989, Schick and Vaughn 1999) encourage people to avoid the use of intuition and instead to base their judgments and decisions on reasoning strategies that are less likely to lead to common errors in reasoning. From this perspective, decision making should be a matter of calculation, not intuition.

While I agree that intuition-based decision making can lead to many problems, I also think that calculation-based decision making of the sort recommended by psychologists and economists has some serious pitfalls. In this chapter, I will try to offer a synthesis and partial reconciliation of intuition and calculation models of decision, using a recently developed theory of emotional coherence (Thagard in press). This theory builds on a previous coherence-based theory of decision making developed in collaboration with Elijah Millgram. Understanding decision making in terms of emotional coherence enables us to appreciate the merits of both intuition and calculation as contributors to effective practical reasoning.

Decision as Intuition

Suppose you are a student trying to decide whether to study (1) an Arts subject such as philosophy or art history in which you have a strong interest or (2) a subject such as economics or computer science which may lead to a more lucrative career. To make this decision intuitively is just to go with the option that is supported by your emotional reactions to the two alternatives. You may have a strongly positive gut feeling toward the more interesting subject along with a strongly negative feeling about the more career-oriented one, or your feelings may be just the opposite. More likely is that you feel positive feelings toward both alternatives, along with accompanying anxiety caused by your inability to see a clearly preferable option. In the end, intuitive decision makers choose an option based on what their emotional reactions tell them is preferable.

There is much to be said for intuitive decision making. One obvious advantage is speed: an emotional reaction can be immediate and lead directly to a decision. If your choice is between chocolate and vanilla ice cream, it would be pointless to spend a lot of time and effort deliberating about the relative advantages and disadvantages of the two flavors. Instead, an emotional reaction such as "chocolate ­yum!" can make for a quick and appropriate decision. Another advantage is that basing your decisions on emotions helps to ensure that the decisions take into account what you really care about. If you are pleased and excited about a possible action, that is a good sign that the action promises to accomplish the goals that are genuinely important to you. Finally, decisions based on emotional intuitions lead directly to action: the positive feeling toward an option will motivate you to carry it out.

But emotion-based intuitive decision making can also have some serious disadvantages. An option may seem emotionally appealing because of failure to consider other available options. Intuition may suggest buying chocolate ice cream only because you have failed to consider a lower-fat alternative that would be a healthier choice. Intuition is also subject to the intense craving that drug addicts call "jonesing". If you are jonesing for cocaine, or for a pizza, or for a Mercedes-Benz convertible, your intuition will tell you to choose what you crave, but only because the craving has emotionally swamped other desires that you will be more aware of when the craving is less intense.

Another problem with intuition is that it may be based on inaccurate or irrelevant information. Suppose you need to decide whom to hire for a job. If you are prejudiced against people of a particular sex, race, or ethnicity, then your intuition will tell you not to hire them, even if they have better qualifications for doing the job well. It is difficult to determine introspectively whether your intuitions derive from reliable and relevant information.

Finally, intuitive reasoning is problematic in group situations where decisions need to be made collectively. If other people disagree with your choices, you cannot simply contend that your intuitions are stronger or better than the intuitions of others. Defending your emotional reactions and attempting to reach a consensus with other people requires a more analytical approach than simply expressing your gut feelings.

Decision as Calculation

Experts on decision making recommend a more systematic and calculating approach. For example, Bazerman (1994, p. 4) says that rational decision making should include the following six steps:

1. Define the problem, characterizing the general purpose of your decision.
2. Identify the criteria, specifying the goals or objectives that you want to be able to accomplish.
3. Weight the criteria, deciding the relative importance of the goals.
4. Generate alternatives, identifying possible courses of action that might accomplish your various goals.
5. Rate each alternative on each criterion, assessing the extent to which each action would accomplish each goal.
6. Compute the optimal decision, evaluating each alternative by multiplying the expected effectiveness of each alternative with respect to a criterion times the weight of the criterion, then adding up the expected value of the alternative with respect to all criteria.

We can then pick the alternative with the highest expected value and make a decision based on calculation, not on subjective emotional reactions. Using slightly different terminology, Russo and Shoemaker (1989, ch. 6) recommend essentially the same kind of decision making process based on multiple weighted factors.

Some students dismiss this kind of process as robot-like, and find it offensive that important decisions in their lives might be made mathematically. A cartoon in the New Yorker (Jan. 10, 2000, p. 74) shows a man sitting at a computer and saying to a woman: "I've done the numbers, and I will marry you." Some decisions, at least, seem inappropriately based on doing the numbers. But is the emotional dismissal of Bazerman's 6-step calculation method justified? We can certainly see some notable advantages of the calculation method over the intuition method. First, it is set up to avoid neglecting relevant alternatives and goals. Second, it makes explicit the consideration of how the various alternatives contribute to the various goals. Third, it puts the decision making process out in the open, enabling it to be carefully reviewed by a particular decision maker and also by others involved in a group decision process.
However, the calculation method of decision making may be more difficult and less effective than decision experts claim. Suppose you are trying to decide between two courses of study, say philosophy versus computer science, and you systematically list all the relevant criteria such as how interesting you find the subjects and how well they fit with your career plans. You then weight the criteria and estimate the extent to which each option satisfies them, and proceed to a calculation of the expected value of the competing choices. Having done this, you find that the expected value of one option, say philosophy, exceeds that of the other. But what if you then have the reaction ­ "I don't want to do that!" Your emotional reaction need not be crazy, because it may be that the numerical weights that you put on your criteria do not reflect what you really care about. Moreover, your estimates about the extent to which different actions accomplish your goals may be very subjective and fluid, so that your unconscious estimation is at least as good as your conscious one. I once knew someone who told me that she made decisions by first flipping a coin, with heads for one option and tails for another. When the coin came up heads, she would note her emotional reaction, which gave her a better idea of whether she really wanted the option associated with heads. She then used this emotional information to help her make a choice between the two options.

There is empirical evidence that calculation may sometimes be inferior to intuition in making good judgments. Damasio (1994) describes people with injuries that have disconnected the parts of their brains that perform verbal reasoning and numerical calculation from emotional centers such as the amygdala. With their abstract reasoning abilities intact, you might think that the patients become paragons of rationality, like Spock or Data in Star Trek. On the contrary, these patients tend to make poor interpersonal decisions. Damasio conjectures that the deficiencies arise because the brain damage prevents the patients from making emotional evaluations that involve somatic markers, bodily states that indicate the positive or negative emotional value of different possibilities. The problem is that the patients just do not know what they care about. Wilson and Schooler (1991) report research that shows that there are domains where people's intuitive judgments may be more effective than their more systematic, deliberative ones. They studied college students' preferences for brands of strawberry jams and for college courses, and found that students who were asked to analyze the reasons for their preferences ended up with choices that corresponded less with expert opinion than did the choices of less analytical students. Wilson and Schooler conjecture that this happens because analyzing reasons can focus people's attention on relatively unimportant criteria. Lieberman (2000) argues that intuitions are often based on unconscious learning processes that can be interfered with by attempts at explicit learning.
It seems, therefore, that we need a model of decision making that is both more psychologically natural and more normatively effective than the calculation model. I will now argue that we can get better accounts both of how decisions are made and of how they should be made by understanding practical inference in terms of emotional coherence.

Decision as Coherence

Decision making is a kind of inference, but what is inference? Many philosophers have taken deductive logic as the model for inference. Here is a sort of deductive practical inference:

Whenever you want ice cream, you should order chocolate.
You want ice cream.
Therefore, you should order chocolate.

Unfortunately, we rarely have general rules that tell us exactly what to do, so deduction is not a good model for practical inference. A second familiar model of inference is calculation, useful for example in solving arithmetical problems and working with probability theory. But there is a third general model of inference that advocates the following rule: Accept a representation if and only if it coheres maximally with the rest of your representations. Many philosophers have advocated coherence theories of inference but have left rather vague how to maximize coherence (e.g. Harman 1986, Brink 1989, and Hurley 1989). A precise and general model of coherence-based inference can be constructed in terms of constraint satisfaction (Thagard and Verbeurgt 1998, Thagard in press).

When we make sense of a text, a picture, a person, or an event, we need to construct an interpretation that fits with the available information better than alternative interpretations. The best interpretation is one that provides the most coherent account of what we want to understand, considering both pieces of information that fit with each other and pieces of information that do not fit with each other. For example, when we meet unusual people, we may consider different combinations of concepts and hypotheses that fit together to make sense of their behavior.
Coherence can be understood in terms of maximal satisfaction of multiple constraints, in a manner informally summarized as follows:

1. Elements are representations such as concepts, propositions, parts of images, goals, actions, and so on.
2. Elements can cohere (fit together) or incohere (resist fitting together). Coherence relations include explanation, deduction, facilitation, association, and so on. Incoherence relations include inconsistency, incompatibility, and negative association.
3. If two elements cohere, there is a positive constraint between them. If two elements incohere, there is a negative constraint between them.
4. Elements are to be divided into ones that are accepted and ones that are rejected.
5. A positive constraint between two elements can be satisfied either by accepting both of the elements or by rejecting both of the elements.
6. A negative constraint between two elements can be satisfied only by accepting one element and rejecting the other.
7. The coherence problem consists of dividing a set of elements into accepted and rejected sets in a way that satisfies the most constraints.

Computing coherence is a matter of maximizing constraint satisfaction, and can be accomplished approximately by several different algorithms. The most psychologically appealing models of coherence optimization are provided by connectionist algorithms. These use neuron-like units to represent elements and excitatory and inhibitory links to represent positive and negative constraints. Settling a connectionist network by spreading activation results in the activation (acceptance) of some units and the deactivation (rejection) of others. Coherence can be measured in terms of the degree of constraint satisfaction accomplished by the various algorithms. In general, the computational problem of exactly maximizing coherence is very difficult, but there are effective algorithms for approximating the maximization of coherence construed as constraint satisfaction (Thagard and Verbeurgt 1998).

I will now make this account of coherence more concrete by showing how it applies to inference about what to do. Elijah Millgram and I have argued that practical inference involves coherence judgments about how to fit together various possible actions and goals; (Millgram and Thagard 1996, Thagard and Millgram 1995). On our account, the elements are actions and goals, the positive constraints are based on facilitation relations (the action of going to Paris facilitates my goal of having fun), and the negative constraints are based on incompatibility relations (you cannot go to Paris and London at the same time). Deciding what to do is based on inference to the most coherent plan, where coherence involves evaluating goals as well as deciding what to do.
More exactly, deliberative coherence can be specified by the following principles:

Principle 1. Symmetry. Coherence and incoherence are symmetrical relations: If a factor (action or goal) F1 coheres with a factor F2, then F2 coheres with F1.
Principle 2. Facilitation. Consider actions A1 ... An that together facilitate the accomplishment of goal G. Then
(a) each Ai coheres with G,
(b) each Ai coheres with each other Aj, and
(c) the greater the number of actions required, the less the coherence among actions and goals.
Principle 3 . Incompatibility.
(a) If two factors cannot both be performed or achieved, then they are strongly incoherent.
(b) If two factors are difficult to perform or achieve together, then they are weakly incoherent.
Principle 4. Goal priority. Some goals are desirable for intrinsic or other non-coherence reasons.
Principle 5. Judgment. Facilitation and competition relations can depend on coherence with judgments about the acceptability of factual beliefs.
Principle 6. Decision. Decisions are made on the basis of an assessment of the overall coherence of a set of actions and goals.

In order to assess overall coherence, we can use the computer program DECO (short for "Deliberative Coherence"). DECO represents each element (goal or action) by a neuron-like unit in an artificial neural network and then spreads activation through the network in a way that activates some units and deactivates others. At the end of the spread of activation, the active units represent elements that are accepted, while the deactivated ones represent elements that are rejected. DECO provides an efficient and usable way to compute the most coherent set of actions and goals.

At first glance, deliberative coherence might seem like a variant of the calculation model of decision making. Figuring out which action best coheres with your goals sounds like Bazerman's calculation of the expected value of alternatives based on the extent to which they satisfy weighted criteria. But there are some crucial differences. Unlike Bazerman's proposal, the deliberative coherence model of decision does not take the weights of the goals as fixed. In DECO, units representing some of the goals get initial activation in accord with principle 4, goal priority, but their impact depends on their relation to other goals: even a basic goal can be deactivated, at least partially, by other goals. The impact of goals on decision making depends on their activation, which depends on their relation to other goals and to various actions. For example, students trying to decide what to do on the weekend might start off thinking that what they most want to do is to have fun, but realize that having fun is not so important because it conflicts with other goals such as studying for an important exam or saving money to pay next term's tuition.

Psychologically, decision as coherence is very different from decision as calculation. Calculations are conscious and explicit, displayable to everyone on pencil and paper. In contrast, if coherence maximization in human brains is similar to what happens in the artificial neural networks used in DECO, then assessment of coherence is a process not accessible to consciousness. What comes to consciousness is only the result of the process of coherence maximization: the realization that a particular action is the one I want to perform. Thus, as an account of how decisions are made by people, deliberative coherence is closer to the intuition model of decision than to the calculation model. Coherence is not maximized by an explicit, consciously accessible calculation, but by an unconscious process whose output is the intuition that one action is preferable to others. There is, however, a major difference between the deliberative coherence account of decision making and the intuition account: intuitions about what to do are usually emotional, involving feelings that one action is a good thing to do and that alternatives are bad things to do. Fortunately, coherence theory can naturally be extended to encompass emotional judgments.

Emotional Coherence

In the theory of coherence stated above, elements have the epistemic status of being accepted or rejected. We can also speak of degree of acceptability, which in artificial neural network models of coherence is interpreted as the degree of activation of the unit that represents the element. I propose that elements in coherence systems have, in addition to acceptability, an emotional valence, which can be positive or negative. Depending on the nature of what the element represents, the valence of an element can indicate likability, desirability, or other positive or negative attitude. For example, the valence of Mother Theresa for most people is highly positive, while the valence of Adolf Hitler is highly negative. Many other researchers have previously proposed introducing emotion into cognitive models by adding valences or affective tags (Bower 1981, 1991; Fiske and Pavelchak 1986; Lodge and Stroh 1993; Ortony, Clore, and Collins 1988; Sears, Huddy, Schaffer 1986). Kahneman (1999) reviews experimental evidence that evaluation on the good/bad dimension is a ubiquitous component of human thinking.

Just as elements are related to each other by the positive and negative deliberative constraints described in the last section, they also can be related by positive and negative valence constraints. Some elements have intrinsic positive and negative valences, for example pleasure and pain. Other elements can acquire valences by virtue of their connections with elements that have intrinsic valences. These connections can be special valence constraints, or they can be any of the constraints posited by the theory of deliberative coherence. For example, if someone has a positive association between the concepts of dentist and pain, where pain has an intrinsic negative valence, then dentist can acquire a negative valence. However, just as the acceptability of an element depends on the acceptability of all the elements that constrain it, so the valence of an element depends on the valences of all the elements that constrain it.

The basic theory of emotional coherence can be summarized in three principles analogous to the qualitative principles of coherence above:

1. Elements have positive or negative valences.
2. Elements can have positive or negative emotional connections to other elements.
3. The valence of an element is determined by the valences and acceptability of all the elements to which it is connected.

As already mentioned, coherence can be computed by a variety of algorithms, but the most psychologically appealing model, and the model that first inspired the theory of coherence as constraint satisfaction, employs artificial neural networks. In this connectionist model, elements are represented by units, which are roughly analogous to neurons or neuronal groups. Positive constraints between elements are represented by symmetric excitatory links between units, and negative constraints between elements are represented by symmetric inhibitory links between units. The degree of acceptability of an element is represented by the activation of a unit, which is determined by the activation of all the units linked to it, taking into account the strength of the various excitatory and inhibitory links.

It is straightforward to expand this kind of model into one that incorporates emotional coherence. In the expanded model, called "HOTCO" for "hot coherence," units have valences as well as activations, and units can have input valences to represent their intrinsic valences. Moreover, valences can spread through the system in a way very similar to the spread of activation, except that valence spread depends in part on activation spread. An emotional decision emerges from the spread of activation and valences through the system because nodes representing some actions receive positive valence while nodes representing other actions receive negative valence. The gut feeling that comes to consciousness is the end result of a complex process of cognitive and emotional constraint satisfaction. Emotional reactions such as happiness, anger, and fear are much more complex than positive and negative valences, so HOTCO is by no means a general model of emotional cognition. But it does capture the general production by emotional inference of positive and negative attitudes toward objects, situations, and choices.

It might seem that we can now abandon the cognitive theory of deliberative coherence for the psychologically richer theory of emotional coherence, but that would be a mistake for two reasons. First, emotional coherence must interconnect with other kinds of coherence that involve inferences about what is acceptable as well as about what is emotionally desirable. The valence of an element does not depend just on the valences of the elements that constrain it, but also on their acceptability. Attaching a negative valence to the concept dentist, if it does not already have a negative valence from previous experience, depends both on the negative valence for causes-pain and the acceptability (confidence) of causes-pain in the current context. The inferential situation here is analogous to expected utility theory, in which the expected utility of an action is calculated by summing, for various outcomes, the result of multiplying the probability of the outcome times the utility of the outcome. The calculated valence of an element is like the expected utility of an action, with degrees of acceptability analogous to probabilities and valences analogous to utilities. There is no reason, however, to expect degrees of acceptability and valences to have the mathematical properties that define probabilities and utilities. Because the valence calculation depends on the acceptability of all the relevant elements, it can be affected by other kinds of coherence. For example, the inference concerning whether to trust someone depends largely on the valence attached to them based on all the information you have about them, where this information derives in part from inferences based on explanatory, analogical, and conceptual coherence (Thagard in press).

The second reason for not completely replacing the cold (nonemotional) theory of deliberative coherence with the hot theory of emotional coherence is that people can sometimes come up with incompatible hot and cold judgments about what to do. Unconsciously using deliberative coherence may produce the judgment that you should not do something, while emotional coherence leads you in a different direction. For example, students seeing the first nice spring day at the end of a long Canadian winter might decide emotionally to go outside and enjoy it, while at the same time reasoning that the alternative of finishing up overdue end-of-term projects is more coherent with their central goals such as graduating from university. I am not the only person capable of thinking: "The best thing for me to do is X, but I'm going to do Y." Jonesing in reaction to vivid stimuli can make emotional coherence swamp deliberative coherence.
The theory of emotional coherence provides a psychologically realistic way of understanding the role of intuition in decisions. My gut feeling that I should go to Paris is the result of an unconscious mental process in which various actions and goals are balanced against each other. The coherence process involves both inferences about what I think is true (e.g. I'll have fun in Paris) and inferences about the extent to which my goals will be accomplished. But the coherence computation determines not only what elements will be accepted and rejected, but also an emotional reaction to the element. It is not just "go to Paris ­ yes" or "go to Paris ­ no", but "go to Paris ­ yeah!" or "go to Paris ­ yuck!".

As we just saw, however, emotional coherence may be better as a descriptive theory of how people make decisions than as a normative theory of how people should make decisions. Judgments based on emotional coherence may be subject to the same criticisms that I made against intuitive decisions: susceptibility to jonesing and to failure to consider the appropriate range of actions and goals. I doubt, however, that people are capable of making decisions without recourse to emotional coherence ­ that is just how our brains are constituted. For normative purposes, therefore, the best course is to adopt procedures that interact with emotional coherence to produce intuitions that are informed and effective.

Using Intuition and Emotion to Make Good Decisions

The theory of emotional coherence shows how people's gut feelings about what to do may sometimes emerge from integrative unconscious judgments about the actions that might best accomplish their goals. But it also applies to cases where people's intuitions are too quick and uninformed. How can students and other people be helped to ensure that their decisions are based on informed intuition?

For important decisions, I recommend that, rather than leaping to an immediate intuitive choice, people should follow a procedure something like the following:

Informed Intuition
1. Set up the decision problem carefully. This requires identifying the goals to be accomplished by your decision and specifying the broad range of possible actions that might accomplish those goals.
2. Reflect on the importance of the different goals. Such reflection will be more emotional and intuitive than just putting a numerical weight on them, but should help you to be more aware of what you care about in the current decision situation. Identify goals whose importance may be exaggerated because of jonesing or other emotional distortions.
3. Examine beliefs about the extent to which various actions would facilitate the different goals. Are these beliefs based on good evidence? If not, revise them.
4. Make your intuitive judgment about the best action to perform, monitoring your emotional reaction to different options. Run your decision past other people to see if it seems reasonable to them.

This procedure combines the strengths and avoids the weaknesses of the intuition and calculation models of decision making. Like the intuition model, it recognizes that decision making is an unconscious process that involves emotions. Like the calculation model, it aims to avoid decision errors caused by unsystematic and unexamined intuitions. One drawback of the Informed Intuition procedure is that it is not so intersubjective as the calculation model, in which the numerical weights and calculations can be laid out on the table for all to see. It would certainly be a useful exercise in many cases for people to go through the steps of producing a calculation in order to provide some information about how different people are seeing the situation. Ultimately, however the individual decision makers will have to make decisions based on their own intuitive judgments about what is the right thing to do. The members of the group may be poor at specifying the emotional weights they put on different goals, and they may be unaware of their assumptions about the extent to which different actions facilitate different goals. Achieving consensus among a group of decision makers may require extensive discussion that reveals the goals and beliefs of decision makers to themselves as well as to others. It is much easier to identify jonesing and other emotional distortions in others than in yourself. The discussion, including the exercise of working through a calculation together, may help the members of the group converge on evaluations of goal importance and belief plausibility that produce a shared reaction of emotional coherence. Scientific consensus concerning competing scientific theories can emerge from a process of individual coherence and interpersonal communication (Thagard 1999, ch. 7), but conflict resolution concerning what to do requires a more complex process of comparing and communicating the diverse goals driving the various decision makers. A crucial part of this process is becoming aware of the emotional states of others, which may benefit as much from face-to-face interactions involving perception of people's physical as from purely verbal communication.

Informed Intuition is a much more complicated process of decision making than the practical syllogism commonly discussed by philosophers. Millgram (1997, p. 41) gives the following example:

1. Delicious things should be eaten. [major premise]
2. This cake is delicious. [minor premise]
3. Eat the cake. [conclusion]

The practical syllogism gives an inadequate picture of decision making, both descriptively and normatively. Descriptively it fails to notice that the decision to eat cake is crucially influenced by the emotional value of the action of eating cake. Normatively it fails to see that deciding is a matter of deliberative coherence which has to balance competing goals (e.g. eat something delicious, be slim, be healthy) and to evaluate competing actions (e.g. eat the cake, eat an apple, drink Perrier). On the coherence model of inference, reasoning and inference are very different. Reasoning is verbal and linear, like the practical syllogism and proofs in formal logic. But inference is an unconscious mental process in which many factors are balanced against each other until a judgment is reached that accepts some beliefs and rejects others in a way that approximately maximizes coherence.

This does not mean that practical and theoretical reasoning should be sneered at. Reasoning is a verbal, conscious process that is easily communicated to other people. People are rarely convinced by an argument directly, but the fact that reasoning does not immediately translate into inference does not make it pointless. Making reasoning explicit in decisions helps to communicate to all the people involved what the relevant goals, actions, and facilitation relations might be. If communication is effective, then the desired result will be that each decision maker will make a better informed intuitive decision about what to do.

Improving inference is both a matter of recognizing good inference procedures such as Informed Intuition and watching out for errors that people commonly make. Such errors are usually called fallacies by philosophers and biases by psychologists. Psychologists, economists, and philosophers have identified a variety of error tendencies in decision making, such as overrating sunk costs, using bad analogies, and being overconfident in judgments. Noticing the role of emotional coherence in decision making enables us to expand this list to include emotional determinants of bad decision making such as jonesing and failing to perceive the emotional attitudes of other people. In this paper I have emphasized the positive strategy of making decisions using a recommended procedure, Informed Intuition, but a fuller account would also develop the negative strategy of avoiding various tendencies that are natural to human thinking and that often lead to poor decisions.

The coherence model of decision making allows goals to be adjusted in importance while evaluating a decision, but it does not address the question of how we adopt new goals. Millgram's (1997) account of practical induction is useful for describing how people in novel situations can develop new interests that provide them with new goals. A full theory of decision making would have to include an account of where human goals come from and how they can be evaluated. People who base their decisions only on the goals of sex, drugs, and rock and roll may achieve local coherence, but they have much to learn about the full range of pursuits that enrich human lives.


I have tried in this paper to provide students and other people with a model of decision making that is both natural and effective. Practical inference is not simply produced by practical syllogisms or cost-benefit calculations, but requires assessment of the coherence of positively and negatively interconnected goals and actions. This assessment is an unconscious process based in part on emotional valences attached to the various goals to be taken into consideration, and yields a conscious judgment that is not just a belief about what is the best action to perform but also a positive emotional attitude toward that action. Reason and emotion need not be in conflict with each other, if the emotional judgment that arises from a coherence assessment takes into account the relevant actions and goals and the relations between them. The procedure I recommend, Informed Intuition, shows how decisions can be both intuitive and reasonable.

This research is supported by the Natural Sciences and Engineering Research Council of Canada. The sections on decision as coherence and emotional coherence contain excerpts from Thagard (in press). I am grateful to Elijah Millgram for comments on an earlier draft. Various papers on coherence can be found on my Web site:


Bazerman, M. H. (1994). Judgment in managerial decision making. New York: John Wiley.
Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129-148.
Bower, G. H. (1991). Mood congruity of social judgments. In J. P. Forgas (Ed.), Emotion and social judgments (pp. 31-53). Oxford: Pergamon Press.
Damasio, A. R. (1994). Descartes' error. New York: G. P. Putnam's Sons.
Fiske, S., & Pavelchak, M. (1986). Category-based vs. piecemeal-based affective responses: Developments in schema-triggered affect. In R. Sorrentino & E. Higgins (Eds.), Handbook of motivation and cognition (Vol. 1, pp. 167-203). New York: Guilford.
Gilovich, T. (1991). How we know what isn't so. New York: Free Press.
Kahneman, D. (1999). Objective happiness. In D. Kahneman, E. Diener, and N. Schwarz (Eds.), Well-being: Foundations of hedonic psychology. New York: Russell Sage Foundation.
Liebermann, M. D. (2000). Intuition: A social cognitive neuroscience approach. Psychological Bulletin, 126: 109-137.
Lodge, M., & Stroh, P. (1993). Inside the mental voting booth: An impression-driven process model of candidate evaluation. In S. Iyengar & W. J. McGuire (Eds.), Explorations in political psychology (pp. 225-295). Durham: Duke University Press.
Millgram, E. (1997). Practical induction. Cambridge, MA: Harvard University Press.
Millgram, E., & Thagard, P. (1996). Deliberative coherence. Synthese, 108, 63-88.
Ortony, A., Clore, G. L., & Collins, A. (1988). The cognitive structure of emotions. Cambridge: Cambridge University Press.
Russo, J. E., & Schoemaker, P. J. H. (1989). Decision traps. New York: Simon & Schuster.
Schick, T., Jr., & Vaughn, L. (1999). How to think about weird things. (second ed.). Mountain View, CA: Mayfield.
Sears, D., Huddy, L., & Schaffer, L. (1986). A schematic variant of symbolic politics theory, as applied to racial and gender equality. In R. Lau & D. Sears (Eds.), Political cognition (pp. 159-202). Hillsdale, NJ: Erlbaum.
Thagard, P. (1999). How scientists explain disease. Princeton: Princeton University Press.
Thagard, P. (in press). Coherence in thought and action. Cambridge, MA: MIT Press.
Thagard, P., & Millgram, E. (1995). Inference to the best plan: A coherence theory of decision. In A. Ram & D. B. Leake (Eds.), Goal-driven learning: (pp. 439-454). Cambridge, MA: MIT Press.
Thagard, P., & Verbeurgt, K. (1998). Coherence as constraint satisfaction. Cognitive Science, 22, 1-24.
Wilson, T. D., & Schooler, J. W. (1991). Thinking too much: Introspection can reduce the quality of preferences and decisions. Journal of Personality and Social Psychology, 60, 181-192.


Back to Paul Thagard's recent articles table of contents.

Back to emotion articles table of contents.

Back to Computational Epistemology Laboratory.