As noted by Newell & Shanks (N&S), recent years have seen numerous authors extolling the virtues of unconscious processes in decision making. From some accounts, it would appear that we would all do better if we were just to “let our unconscious do the thinking.”
N&S question several claims for the influence of unconscious influences, and their findings give pause for solid conscious thought. What precisely are the influences of unconscious processes in decision making? How can these be reliably detected? These are difficult questions, and my goal here is to add some elements to the debate.
My first point is to express surprise that N&S do not explore more the topic of automatic processes and the extent to which these do or do not involve unconscious influences. It is clear that in learning physical skills (imagine riding a bicycle), people go through a process whereby, at the outset, they are conscious of their movements, but once skilled, they can control what they do without paying much, if any, conscious attention. It is difficult to prove or disprove the role of unconscious influences in these kinds of situations, but an analogy can clearly be made with judgments or decisions. Is the use of an automatic process conscious or, on the contrary, an unconscious reaction to a process that has already been initiated?
One interesting phenomenon lies in decisions that have been taken automatically but for which there is no longer a conscious memory. (Does this mean they were subject to unconscious influences?) For example, like many other people I am supposed to swallow a pill every morning. Occasionally, I cannot remember if I have done this. Should I then deliberately take – perhaps – another pill? Fortunately, my pill container indicates days of the week. Thus, when I wonder whether I have taken my pill, I look to see if the pill for that day is still there. If it is, I didn't take my pill; if it isn't, I did. Parenthetically, this example does not satisfy N&S's criteria for detecting an unconscious influence because, if I had been asked just after taking the pill whether or not I had done so, my answer would undoubtedly be that I had – it is the delay that appears to erase this automatic event from memory.
A second topic that N&S do not highlight is the role of emotions in decision making and whether this involves unconscious influences. There is a considerable literature that illustrates how emotional states affect judgments of risk and even risky decision making (Andrade & Ariely Reference Andrade and Ariely2009; Slovic & Peters Reference Slovic and Peters2006). For example, in one research program, my colleagues and I used experience sampling to collect judgments of mood and emotion as well as assessments of risk (Hogarth et al. Reference Hogarth, Portell, Cuxart and Kolev2011). We found that emotional states explained variability in risk judgments over and above rational factors of probabilities and magnitudes of potential losses. We did not attempt to determine whether participants were aware that their emotional state was impacting their risk judgments. However, from an N&S perspective, future studies could clearly do this. My hypothesis is that people are not always aware of how emotions influence their decisions in the same way that these might be affected by, for example, relative states of hunger (Danziger et al. 2011).
A third point deals with a difficulty in interpreting the differences that researchers and participants in experiments have concerning whether a variable has had a causal influence on a decision. An example given in a seminar some 30 years ago by Richard Nisbett illustrates the point. (Incidentally, although I like and remember the scenario, I do not recall the specific point that Nisbett was illustrating!)
Imagine that a social psychologist is conducting a study on the influence of lighting in restaurants on romantic attachment. Couples are recruited for blind dates involving a meal at a restaurant, and there are two experimental conditions to which couples are randomly assigned. In one, the restaurant is fully illuminated; in the other, the lights have been dimmed. The dependent variable is the proportion of couples who decide to meet again after the meal. Now imagine that this variable is significantly greater for couples in the dimmed lighting condition. What does this mean? For the social psychologist, the inference is that dimmed lighting fosters romantic attachment. After all, this was the variable that was manipulated experimentally, and there was an effect.
Now imagine that you ask the couples whether the lighting in the restaurant influenced their decisions to meet again. Almost certainly, they would deny that this had played any role.
So who is “correct” – the social psychologist or the couples? It is possible to make an argument that both are correct. What differs between the two is the definition of the causal background – or “field” – against which the causal inference is made (Einhorn & Hogarth Reference Einhorn and Hogarth1986; Mackie Reference Mackie1965; Reference Mackie1974). For the social psychologist, the causal field involves both experimental groups (with and without dimmed lighting), and the difference in the levels of lighting is a “difference-in-the-field” and thus a potential causal factor. The causal fields of the experimental participants, however, contain no such difference. The experience for each couple consists entirely of dimmed or full lighting, and they never experience the differences between the two conditions. For the couples, therefore, there is no way that they can assign cause to the level of lighting. For each couple, lighting is a constant and thus not causally relevant.
Advocates of influences of unconscious effects on decisions would undoubtedly agree with the social psychologist. However, this conclusion only holds at one level of analysis (i.e., causal field). In general, we should be clear at which levels we wish to draw conclusions.
As noted by Newell & Shanks (N&S), recent years have seen numerous authors extolling the virtues of unconscious processes in decision making. From some accounts, it would appear that we would all do better if we were just to “let our unconscious do the thinking.”
N&S question several claims for the influence of unconscious influences, and their findings give pause for solid conscious thought. What precisely are the influences of unconscious processes in decision making? How can these be reliably detected? These are difficult questions, and my goal here is to add some elements to the debate.
My first point is to express surprise that N&S do not explore more the topic of automatic processes and the extent to which these do or do not involve unconscious influences. It is clear that in learning physical skills (imagine riding a bicycle), people go through a process whereby, at the outset, they are conscious of their movements, but once skilled, they can control what they do without paying much, if any, conscious attention. It is difficult to prove or disprove the role of unconscious influences in these kinds of situations, but an analogy can clearly be made with judgments or decisions. Is the use of an automatic process conscious or, on the contrary, an unconscious reaction to a process that has already been initiated?
One interesting phenomenon lies in decisions that have been taken automatically but for which there is no longer a conscious memory. (Does this mean they were subject to unconscious influences?) For example, like many other people I am supposed to swallow a pill every morning. Occasionally, I cannot remember if I have done this. Should I then deliberately take – perhaps – another pill? Fortunately, my pill container indicates days of the week. Thus, when I wonder whether I have taken my pill, I look to see if the pill for that day is still there. If it is, I didn't take my pill; if it isn't, I did. Parenthetically, this example does not satisfy N&S's criteria for detecting an unconscious influence because, if I had been asked just after taking the pill whether or not I had done so, my answer would undoubtedly be that I had – it is the delay that appears to erase this automatic event from memory.
A second topic that N&S do not highlight is the role of emotions in decision making and whether this involves unconscious influences. There is a considerable literature that illustrates how emotional states affect judgments of risk and even risky decision making (Andrade & Ariely Reference Andrade and Ariely2009; Slovic & Peters Reference Slovic and Peters2006). For example, in one research program, my colleagues and I used experience sampling to collect judgments of mood and emotion as well as assessments of risk (Hogarth et al. Reference Hogarth, Portell, Cuxart and Kolev2011). We found that emotional states explained variability in risk judgments over and above rational factors of probabilities and magnitudes of potential losses. We did not attempt to determine whether participants were aware that their emotional state was impacting their risk judgments. However, from an N&S perspective, future studies could clearly do this. My hypothesis is that people are not always aware of how emotions influence their decisions in the same way that these might be affected by, for example, relative states of hunger (Danziger et al. 2011).
A third point deals with a difficulty in interpreting the differences that researchers and participants in experiments have concerning whether a variable has had a causal influence on a decision. An example given in a seminar some 30 years ago by Richard Nisbett illustrates the point. (Incidentally, although I like and remember the scenario, I do not recall the specific point that Nisbett was illustrating!)
Imagine that a social psychologist is conducting a study on the influence of lighting in restaurants on romantic attachment. Couples are recruited for blind dates involving a meal at a restaurant, and there are two experimental conditions to which couples are randomly assigned. In one, the restaurant is fully illuminated; in the other, the lights have been dimmed. The dependent variable is the proportion of couples who decide to meet again after the meal. Now imagine that this variable is significantly greater for couples in the dimmed lighting condition. What does this mean? For the social psychologist, the inference is that dimmed lighting fosters romantic attachment. After all, this was the variable that was manipulated experimentally, and there was an effect.
Now imagine that you ask the couples whether the lighting in the restaurant influenced their decisions to meet again. Almost certainly, they would deny that this had played any role.
So who is “correct” – the social psychologist or the couples? It is possible to make an argument that both are correct. What differs between the two is the definition of the causal background – or “field” – against which the causal inference is made (Einhorn & Hogarth Reference Einhorn and Hogarth1986; Mackie Reference Mackie1965; Reference Mackie1974). For the social psychologist, the causal field involves both experimental groups (with and without dimmed lighting), and the difference in the levels of lighting is a “difference-in-the-field” and thus a potential causal factor. The causal fields of the experimental participants, however, contain no such difference. The experience for each couple consists entirely of dimmed or full lighting, and they never experience the differences between the two conditions. For the couples, therefore, there is no way that they can assign cause to the level of lighting. For each couple, lighting is a constant and thus not causally relevant.
Advocates of influences of unconscious effects on decisions would undoubtedly agree with the social psychologist. However, this conclusion only holds at one level of analysis (i.e., causal field). In general, we should be clear at which levels we wish to draw conclusions.