Knowledge miscalibration and its origins
Miscalibration of knowledge is directional and is often manifested as overconfidence (Alba and Hutchinson 1987; Fischhoff, Slovic, and Lichtenstein 1977). People tend to overestimate— rather than underestimate—their knowledge, and findings from several different research areas suggest that overconfidence is common.
We do not mean to suggest that all overconfidence is driven by miscalibration of knowledge.A boxer who has been training more intensely than ever before may believe that he is a better boxer than he really is, and that he is better than others, but his overconfidence has little to do with knowledge of boxing. Moore and Healy (2008) categorize overconfidence into three types: overestimation, overplacement, and overprecision. Because knowledge miscalibration can take all three forms, we believe it is one of the most important drivers of overconfidence.Take, for example, someone who believes they know more about financial investing than they really do.They may exhibit all three types of overconfidence by thinking they are better at choosing investments than they really are (overestimation), by thinking they are better than others at investing (overplacement), and by believing that their investment picks have a higher probability of success than they do in reality (overprecision).1Rozenblit and Keil (2002) demonstrated that people systematically overestimate their knowledge of even simple things, suffering from an “illusion of explanatory depth” (IOED). In these studies, subjects were first asked to evaluate how well they understood simple mechanical objects such as a zipper or toilet, on a 1-to-7 scale. In the next step, the experimenters asked subjects to write step-by-step causal explanations for how these items work. Finally, subjects were asked to re-rate their understanding. The result was that they significantly reduced their reported levels of understanding compared to their initial levels.
The difficulty of the step-by-step explanation task made it salient to subjects that their initial mental models were impoverished and overly simple. Separate research on judgment and decision making has used the difference between people's average confidence ratings for their answers to fact-based questions and their number of correct answers to demonstrate that people are overconfident (Fischhoff et al., 1977; Yates, Lee, and Bush, 1997;Yates, Lee, and Shinotsuka, 1996).There are several plausible explanations for why knowledge overconfidence is pervasive. First, it may be socially adaptive to be confident. Confident people are perceived to be more competent by their peers, and achieve higher status in their groups (Anderson, Brion, Moore, and Kennedy, 2012;Tetlock and Gardner, 2015).When evaluating others, people tend to use a “confidence heuristic,” inferring that those with more confidence possess greater knowledge, and liking experts who project more certainty (Gaertig and Simmons, 2018; Price and Stone, 2004). Not only is confidence socially adaptive—it may be evolutionarily adaptive.An individual's confidence in their social status enables positive risk-taking such as seeking necessary support from others in response to social negativity (Baumeister, Heatherton, and Tice, 1993; Park and Maner, 2009). Self-confidence can also signal social dominance, which is sexually attractive to potential mates (Bernstein, 1980; Ellis, 1995).
We suggest that another major contributing factor to overconfidence is the way in which people mentally represent and store knowledge. In Rozenblit and Keil's (2002) IOED studies overestimation of knowledge appears to have been driven by a neglect of underlying complexity. Subjects reduced their reported levels of understanding compared to their initial levels because the difficulty of the step-by-step explanation task made it salient that the domains or objects in the experiment were much more complex than they originally thought.
The IOED is also consistent with separate work arguing that knowledge overconfidence is driven by a tendency to focus on known information and neglect unknowns (Walters, Fernbach, Fox, and Sloman, 2017). Importantly, this miscalibration is particularly dramatic for knowledge of explanations, which is distinguished by a structure of more causal complexity compared to knowledge of facts, procedures, or narratives (Keil, Rozenblit, and Mills, 2004). This is important to note for the current discussion because topics on which there is heated discourse (and a lack of intellectual humility) are frequently those with high levels of complexity, such as the causes and consequences of climate change, changes in the U.S. healthcare system, vaccination, and immigration policy, for example.If knowledge miscalibration is at least partially driven by underestimation of complexity, it is useful to consider why humans systematically underestimate complexity. Although we may prefer to think otherwise, relative ignorance is a default human state; the world is so complex that no individual can truly understand very much of it (S. A. Sloman and Fernbach, 2017). Trying to understand and remember millions of causally inter-related details, processes, and potential outcomes is impossible because the human mind evolved to act effectively in spite of pervasive ignorance (Keil, 2006; S.A. Sloman and Fernbach, 2017).A common way humans cope with the overwhelming complexity of the world is to outsource knowledge to others (Rabb, Fernbach, and Sloman, 2019;Wegner, 2011).They frequently underestimate complexity because most people function normally without ever truly having to deal with it.
This shared nature of human knowledge makes it difficult for people to separate their own knowledge from what their communities know and believe (S. A. Sloman and Fernbach, 2017; S. A. Sloman and Rabb, 2016).This introduces a barrier to intellectual humility because people overestimate their understanding by virtue of simply being part of a community, and get a lift in their sense of understanding just from the knowledge that others have some understanding.
For example, merely telling people that scientists have explained a newly discovered natural phe- nomenon—without saying anything about what the explanation is—causes people to increase their assessment of their own understanding (S. A. Sloman and Rabb, 2016). People even get a boost in their perceived understanding and think they have more knowledge in their heads when they have access to information on the internet (Fisher, Goddu, and Keil, 2015; Ward, n.d.), and fail to remember more information when they think it is being saved by a computer (Sparrow, Liu, and Wegner, 2011). Because people outsource knowledge and fail to recognize it, their mental models are often oversimplified and incorrect. When individuals fail to adequately check their understanding, unjustified attitudes may become entrenched in a community, perpetuating a dangerous cycle of ignorance.35.2