By Eric Vandenbroeck and co-workers
We all have biases, whereby one of the reasons is how we have been socialized. Microsoft’s online UB training, also available publicly, includes videos depicting various everyday workplace scenarios. In one, the only woman on a team tries to add her views and is interrupted multiple times until another member finally notices and asks her to speak.
Yet bias occurs across a wide variety of judgment domains. People in all demographic groups display it, and it is exhibited even by expert reasoners, the highly educated, and the highly intelligent. It has been demonstrated in research studies across various disciplines, including as we have seen in psychology and psychiatry, political science, behavioral economics, legal studies, cognitive neuroscience, and the informal reasoning literature. Bias has been found to occur in every stage of information processing. Studies have shown a tendency toward the biased search for evidence, biased evaluation of evidence, biased assimilation of evidence, biased memory of outcomes, and biased evidence generation.
Our problem is not that we cannot value and respect truth and facts, but that we cannot agree on commonly accepted truth and facts. We believe that our side knows the truth. Post-truth? That describes the other side. The inevitable result is political polarization. However, today science can tell us about confirmation bias: how common it is, how to avoid it, and what purposes it serves.
So, for example, models focusing on the properties of acquired beliefs rather than cognitive processes provide better frameworks for studying confirmation bias.
And because confirmation bias is not predictable from traditional psychological measures, we next will explain how it creates a true blind spot among, for example also cognitive elites. Cognitive elites (those high in intelligence, executive functioning, or other valued psychological dispositions) often predict that they themselves are less biased than other people when queried about other well-known psychological biases (overconfidence bias, omission bias, hindsight bias, anchoring bias).
Noting that researchers in several political and social psychology areas have found content factors to be more potent predictors than individual difference variables. Confirmation bias might be another area of psychology where the properties of the beliefs themselves are more predictive than the individual psychological characteristics of the subjects (such as intelligence or open-mindedness) who hold those beliefs. Studying confirmation bias is important because the tragedy of the communications commons arises when both sides in public policy debates that could be decided by evidence process new information with a confirmation bias. In some cases, when people project prior beliefs that have been arrived at by properly accommodating previous evidence, then some degree of projecting the prior probability, a local confirmation bias, onto new evidence is normatively justified. When we lack previous evidence on the issue in question, we should use the principle of indifference and set our prior probability at .50, which will not influence our evaluation of the new evidence. Instead, what most of us tend to do in this situation is assess how the proposition in question relates to some distal belief of ours, such as our ideology, set, and then project this prior probability onto our evaluation of the new evidence. This is how our society ends up with political partisans on both sides of any issue, seemingly unable to agree on the facts of that issue and never to reach Bayesian convergence. We can at least mitigate, if not remedy, the tragedy of the communications commons by rethinking how we relate to our beliefs. First, it would help if we could realize that we have thought our way to our beliefs much less often than we may have imagined, that the distal beliefs we hold are largely a function of our social learning within the valued groups to which we belong and of our innate propensities to be attracted to certain types of ideas. Dual-inheritance theories of culture have stressed for some time that most people feel in control of their culture and believe they came by most of it by choice. But the truth is, we often have much less choice than we think. We treat our beliefs as possessions when we think that we have thought our way to serve us. The meme’s-eye view leads us to question whether we have thought our way to these beliefs and serve our personal ends. Our memes want to replicate whether they are good for us or not, and they don’t care how they get into us, whether they get in through our conscious thoughts or are simply an unconscious fit to our innate psychological dispositions. The focus of memetics on the properties of beliefs rather than the psychological characteristics of those who hold these beliefs also seems to be consistent with research showing that the degree of confirmation bias is better predicted by the former than by the latter. We are often in situations where we have to calibrate the reasoning of others. This calibration often involves judging the degree of confirmation bias they exhibit, one of the trickiest judgments we have to make.
Subjects are scoring higher on rational thinking dispositions scales and are better able to avoid most biases. This is not true in confirmation bias, where it is the strength of the belief itself rather than the cognitive sophistication of the person holding the belief that predicts the level of confirmation bias. This situation presents a particular obstacle for cognitive elites when evaluating confirmation bias. Their assumption that they are less biased than other people is actually correct for most biases in the heuristics and biases literature. However, this assumption does not hold for confirmation bias, which contributes to our currently vexing partisan political standoffs, both in terms of what knowledge is acquired and how that knowledge is acquired; there is no strong evidence that the Trump voters were more epistemically irrational than the Clinton voters were.
There is no strong support in the empirical literature for attributing a unique problem of rationality to Trump voters. Those who do not find this conclusion palatable might object that the analysis so far seems too narrow. We have discussed that to think rationally; a person needs to take appropriate action given the person’s goals and beliefs and hold beliefs congruent with the available evidence. This third consideration moves us, critically, from a narrow conception of instrumental rationality to a broad one. Traditional views of instrumental rationality are narrow theories because a person’s goals and beliefs are accepted as they are, and evaluation centers only on whether a person optimally satisfies their desires.
It might seem that a narrow conception of rationality that fails to evaluate desires allows a great deal of bad thinking to escape evaluation. But most work in cognitive science disproportionately addresses this narrow form of rationality for a good reason. Broad theories implicate some of the most difficult and vexing issues in philosophy, such as “When is it rational to be (narrowly) rational?” and “What goals are rational to pursue?” Our earlier discussion, in fact, strayed into the territory of broad rationality with my Ted Cruz versus Al Sharpton thought experiment. With that example, I was trying to illustrate the difficulty of evaluating goals. From the standpoint of a voter with the global and groups worldview, the choice was between a candidate who shared the voter’s worldview but whose temperament was poorly suited to the presidency (Sharpton) and a candidate whose worldview was unpalatable to the voter but whose temperament was much better suited to the presidency (Cruz). The point was not to show that one or the other choice was correct for this voter, but to illustrate the difficulty of this type of trade-off and to provoke some associated recognition of the fact that the voter with a citizen and country worldview was presented with a similarly difficult trade-off when faced with a Trump versus Clinton choice. It highlighted the potential confirmation bias involved in such judgments: a Democrat feeling the attraction of Sharpton over Cruz should similarly understand the attraction of Trump over Clinton for Republicans.
Of course, the thought experiment does not depend on a one-to-one feature analogy between Trump versus Clinton and Cruz versus Sharpton, only a gross similarity in the feature trade-off (worldview versus fitness for office). It is also revealing of the confirmation bias operating when we attempt to evaluate goals.
Also, recent developments within the university are making it harder to teach students the decoupling and decontextualizing skills necessary for avoiding confirmation bias.
Why is cognitive decoupling central to avoiding confirmation bias? Two critical functions are enabled by decoupling: inhibition and sustained simulation. The first function, suppression of the automatic response, is akin to the inhibitory processes studied in the executive functioning literature. When we reason hypothetically, we create temporary world models and test out actions in that simulated world. Decoupling our representations from the world enables us to do this. Dealing with these "secondary representations,” keeping them decoupled, is costly in terms of cognitive capacity. However, the tendency to initiate such decoupling for simulation is a dispositional variable, separable from cognitive capacity. This tendency can be developed through experience and training.
Many different theorists have emphasized the importance of decontextualizing in the development of higher-level thought. Thus Jean Piaget’s (1972) conceptualization of formal operational thought places the mechanisms of decontextualizing in positions of paramount importance, and many scholars in the critical thinking literature have emphasized the decontextualizing modes of decentering, detaching, and depersonalizing as the foundational skills of rational thought. Looming large in that literature is the ability to adopt perspectives other than one’s own. The avoidance of confirmation bias is dependent on these perspective-switching abilities and tendencies.
But our ability to switch perspectives will be limited because our human brains are cognitive misers— their basic tendency is to default to processing mechanisms of low computational expense. This is a well-established theme throughout the past fifty years of research in psychology and cognitive science. Miserly cognitive processing arose for sound evolutionary reasons of computational efficiency. Still, that same efficiency will guarantee that perspective switching (to avoid confirmation bias) will not be the default processing action. As we have known for some time, it is cognitively demanding to process information from another person's perspective. Thus we must practice the perspective switching needed to avoid confirmation bias until it becomes habitual. But identity politics prevents this from happening by locking in an automatized group perspective, contextualizing based on preapproved group stances, and viewing perspective switching through decoupling as a sellout to the hegemonic patriarchy.
True perspective switching, reframing that allows us to conceptualize the world in new ways, requires alienation from the self. It requires that we sometimes avoid framing things from the easiest perspective to model, inevitably our own and our most important affinity groups. However, university undergraduates are precisely the stage of life as young adults who need to learn other framing strategies.
Perspective switching is a type of cognitive broccoli. Taking students out of the comfort zone of their identities or those of their tribes was once seen as one of the key purposes of university education. But when the university affirms students in identities they had assumed before they arrived on campus, it is hard to see the value added by the university anymore. In stressing identity politics, the university is simply cheerleading for ice cream. Instead, students need to be taught that the benefits of leaving the comfort and safety of perspectives they have long held are worth the risks, that, in the long run, confirmation processing will never lead them to a deep understanding of the world in which they live.
If we are ever to remedy the tragedy of the science communications commons, if we are ever to have a society that can converge on the truth about important social and public policy issues, we must have institutions that foster decoupling by discouraging the projection of convictions onto evidence. Although they once served that purpose, universities are now the primary incubators and purveyors of intellectually deadening identity politics, whose thinking styles have spread widely into the corporate world in recent years. The James Damore incident at Google is a primary case; an employee was fired for circulating an essay that contained fair comments based on largely accurate social science findings on sex differences.
Members of the identity-politics left have succeeded in making certain research conclusions within the university verboten and making it very hard for any university professor to publish and promote any conclusions they dislike. Faculty now self-censor on a range of topics. The identity politics ideologues have won the on-campus battle to suppress views that they do not like. But they have made the public rightly skeptical about any conclusions that now come out of universities on charged topics, even conclusions that are congenial to the political positions that the ideologues advocate.
Rather than inculcating specific beliefs, open inquiry used to be the sine qua non of the university. With the advent of diversity statements, the goal seems to be tribal: requiring allegiance to specific political content on faculty and students. If the state universities do not require diversity statements, state legislatures should withhold funds from them until they do. Although administrators and faculty organizations may view my recommendation as an attack on the institution as a whole, it is not. Rather, compelling state universities to do away with these statements and, one hopes, persuading private universities to follow their example would represent a valid attempt by the public to steer all universities back to their true mission. Only then will the universities staunch the confirmation bias that is ruining our public communications commons.
Once we have decided on a partisan side, we tend to allow party elites to bundle specific issue positions for us. In many cases, we have given no thought to the issue beyond knowing that our party endorses a certain stance toward it.
And of course, it does not primarily show up at Universities; another example for sure is the media. The selective exposure problems surrounding media entities like Fox News have increased, as other networks (CNN, MSNBC) and traditional media entities such as the New York Times (particularly after the 2016 election) adopt or imitate its business model.
Many true scientists in sociology are demoralized because their field has transitioned from being a social science into a social movement. Soon, granting agencies will become more aware of the ideological bias of state legislatures and taxpayers funding their state universities. This shows that they employe not to foster intellectual diversity but instead often focus on demographic groups.