Politics matter more than credentials in laypeople’s judgments of expertise

Mertcan Güngör; Nathan Ballantyne; Jared B Celniker

doi:10.1038/s41598-026-40053-0

. 2026 Mar 9;16:12765. doi: 10.1038/s41598-026-40053-0

Politics matter more than credentials in laypeople’s judgments of expertise

Mertcan Güngör ^1,^✉, Nathan Ballantyne ², Jared B Celniker ²

PMCID: PMC13096159 PMID: 41803298

Abstract

We explored the cues laypeople use to determine who is an expert on a given topic. In Study 1 (n = 208), we collected participants’ self-reports of cues they consider important when evaluating a putative expert. Across various topics, participants rated good cues (e.g., relevant degree, more experience) as more important for expertise while disregarding poor or superficial cues (e.g., appearance). In Study 2 (n = 498), we conducted a conjoint experiment where we manipulated several cues of expertise in biographies of hypothetical experts. In line with the self-reports, good cues had the strongest effects on trust. In Study 3 (n = 1,776), we examined whether participants would still recognize good cues when they received information about the expert’s political views. Using a highly politicized topic—abortion—we manipulated the expert’s views along with their credentials. Participants trusted likeminded experts more, and the effect of the expert’s views on participants’ trust was more than twice as large as the effect of credentials. Our results paint a complicated picture, where laypeople know what qualities to look for in experts but attend to political information more readily than to good cues of expertise.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-40053-0.

Subject terms: Cultural and media studies, Cultural and media studies, Psychology, Psychology

Introduction

Faced with uncertainty about important questions, some people attempt to “do their own research”^1,2. But knowing nearly anything interesting about the world requires depending on knowledgeable others. For millennia, humans have used division of labor and specialization to construct knowledge beyond what any person working alone could piece together over the course of their lifetime. When we are uncertain about a question but others know the answer, we benefit from the ability to identify the people in the know. We can call such people “experts”.

Identifying experts is an essential task but is easier said than done. The very idea of expertise is in the crosshairs in social and political conflicts. For better or worse, the idea of expertise is connected to the politicization of policy-relevant science as well as the distrust of institutions and elites. In fact, many recent observers have declared a “crisis of expertise”. That “crisis” has been taken to mean many things, including a decline of trust in established authorities and the emergence of “thought leaders”³ and “science influencers”⁴, the existence of rival claims to expertise without a settled way to decide amongst claimants⁵, the use of doubt-inducing strategies by corporations and elites to obscure scientific consensus from the general public⁶, and even the widespread public rejection of any meaningful distinction between the knowledgeable and the ignorant⁷.

Just like talk, people claiming expertise is cheap; we should be shrewd when evaluating such claims. But are there effective strategies for identifying experts? Researchers have noted that strategies are different for experts than for novices or laypeople. Experts, on the one hand, can calibrate others’ expertise by reference to their own—that is, they can leverage their knowledge to identify the knowledgeable^8,9. On the other hand, novices, by definition, do not have expert-level knowledge. A minority view^10,11 holds that novices are necessarily “blind” when trying to find experts and so are forced to exercise trust. The majority view, however, is that novices are not invariably “blind”. Novices can sometimes acquire special types of empirical evidence to help them identify expertise and even reasonably distinguish between rival experts.

Although there is no canonical view about what evidence novices should pay attention to when seeking expertise, theorists agree on some main ideas^9,12–14. For instance, one standard idea is that experts’ experience—perhaps represented in the form of a “track record”—can be a cue for novices to recognize expertise. Another widely noted cue is credentials or certifications. And another cue is the opinions of known experts about a putative expert in question. Of course, not every type of evidence about experts happens to be good evidence—that is, the type of evidence that guides people to accurate judgment concerning the quality of someone’s expertise. Naturally, people can learn facts about experts’ height, weight, sex, ethnicity, and so on, but in many domains such facts are presumably poorly correlated to the quality of experts’ knowledge. Accordingly, such evidence typically constitutes a poor or misleading cue for the identification of expertise. We will rely on that intuitive, rough-and-ready distinction between good cues and poor cues in this article.

Can people attend to good cues when identifying and assessing expertise? There is little existing empirical evidence, but we have strong reason to believe that people can sometimes track good cues. Across many domains of everyday life, we expect people to behave reasonably when assessing expertise. Laypeople, including children, often trust the opinions of individuals who have more relevant expertise^15–19—we are not surprised when a sick person goes to a doctor instead of a plumber. Laypeople often show skepticism toward implausible claims, even when the source of the claims has relevant expertise or holds a position of authority²⁰. That broad kind of reasonableness should be unsurprising, given that people are generally epistemically vigilant²¹.

Of course, being reasonable sometimes doesn’t imply that people assess expertise perfectly. People might implicitly or explicitly identify cues that they sincerely believe are important but are in fact irrelevant for identifying expertise. For instance, people sometimes treat race and gender as indicative of scientists’ credibility²². In other cases, what are good cues in one situation can mislead elsewhere. For instance, having personal experience can be a good cue in most cases¹⁷, but one person’s experience may not be generalizable for topics where desired outcomes highly depend on individual variability (e.g., nutrition, skincare) or luck (e.g., stock market²³. Another example is the phenomenon of “epistemic trespassing”²⁴, which involves experts who stray from their actual domain of competence but nevertheless pass judgment. One case of epistemic trespassing features Linus Pauling, the Nobel Prize-winning chemist who later touted the power of megadoses of vitamin C to treat cancer. When a trespassing expert has impressive credentials or experience in a different domain, their background might easily mislead novices. Despite these imperfections, such errors in lay evaluation of expertise might be correctable through better education and more responsible behavior on the part of experts.

However, there’s reason to suspect that lay evaluations of expertise are subject to biases that are more difficult to overcome. Political partisans evaluate politically-tinged information in biased ways²⁵—and we should expect that their judgments of expertise will be no different. Though some political biases might be more expressive than genuinely believed^26–28, partisan reactions to the COVID-19 pandemic suggest that such biases are often sincere and costly. Despite substantial consensus amongst knowledgeable experts, Republicans trusted public health guidance less and died more because of it²⁹. Many Republicans placed their trust in authorities with epistemologically weaker credentials, coinciding with the increase of many Americans treating “influencers” such as Joe Rogan as news sources³⁰. While the response to public health guidance during the pandemic seems to paint a rosier picture of liberals’ evaluations of expertise, controlled experiments generally show that liberals and conservatives both display a similar tendency to evaluate scientific information in biased ways^31–33. If partisans are prone to biased evaluations of expertise—not mere errors—then different interventions will be required to improve their expertise identification.

The current studies aim to address key empirical gaps in our understanding of how laypeople evaluate expertise. We examined the extent to which people weigh good and poor cues when deciding which expert to trust, and we tested whether their evaluations of expertise are skewed by political biases.

Results

Study 1

In Study 1, we asked participants (n = 208) which factors they thought were relevant when judging whether someone has expertise (Fig. 1). We asked about three politicized topics (abortion, homelessness, police brutality) and three non-politicized ones (nutrition, skincare, stock investments). Attributes that were rated by participants as the most relevant could be classified as “soft skills”—such as personal experience and intelligence—along with those that signal trustworthiness, such as moral character, financial incentives, and reputation among the general public. Those attributes were followed by domain-specific credentials that can be displayed more easily, such as years of education, having a degree relevant to the topic, and peer recognition in the form of awards or prizes.

Participants perceived some attributes as more relevant for expertise depending on the topic. For example, they placed greater importance on the expert’s political preferences when they perceived a topic as more politicized (b = 0.20, SE = 0.04, p <.001) and had stronger moral convictions about it (b = 0.11, SE = 0.04, p =.004). Liberals seemed to care slightly more about the expert’s politics but the relationship was not significant (b = −0.05, SE = 0.03, p =.069). In addition, some attributes were seen as relevant only for one topic but trivial for all others. For example, an expert’s wealth only mattered when getting advice on stock investments (F(5, 463.24) = 61.66, η²_P = 0.40, p <.001); the expert’s gender mattered for abortion (F(5, 471.97) = 55.15, η²_P = 0.37, p <.001); and the expert’s looks mattered for skincare (F(5, 455.99) = 39.36, η²_P = 0.30, p <.001).

Study 2

Participants in our first study identified certain attributes of experts that they found relevant when determining who to trust. In Study 2, we tested whether the self-reported attributes actually impacted people’s trust toward experts using a conjoint experiment. Conjoint experiments are a tool for examining how people simultaneously evaluate multiple attributes that can introduce trade-offs (e.g., demographic characteristics vs. policy preferences of political candidates). Each factor is manipulated simultaneously for each participant, which allows researchers to capture the overall effect of each attribute by calculating the average marginal component effects (AMCEs)^34,35. Unlike full factorial designs, conjoint experiments remain feasible (in terms of sample size and statistical power) as the number of factors increases. In Study 2, we manipulated six attributes that received moderate-to-high relevance ratings in Study 1, along with the expert’s gender, making a conjoint design more practical.

Participants (n = 498) read short biographies of hypothetical experts on the same six topics as the previous study. The expert either had a relevant or an irrelevant major, studied the topic for a long or a short time, studied at an Ivy League institution or a less famous one (e.g., Mayville State University), received many citations and awards from their peers or not, had recognition from the public or not, and had some anecdotal experience with the topic or not. Each attribute was independently randomized, meaning that the experts could be stronger in some respects but weaker for others. We measured the experts’ perceived trustworthiness, expertise, and credibility (see Methods). Even though the former two have been conceptualized as distinct constructs^36,37, we aggregated the items due to the high degree of convergence among them (with α’s between 0.95 and 0.97 across topics); we will refer to the aggregate as “trust” for the sake of simplicity.

We found that people trusted an expert more when the expert had researched the topic for a longer time, had a degree that was relevant to the topic, or received more recognition from the scientific community (Fig. 2). Those three attributes, which we would ordinarily think of as good cues of an expert’s knowledge, had the largest impact on trust across all topics (an increase of 0.54, 0.53, and 0.41 SDs, respectively). But we also found that other cues that are weaker signals of knowledge nevertheless influenced trust. People trusted an expert more if they had some anecdotal experience on the topic, a degree from an Ivy League institution, or many followers of social media. The impact of these poor cues was smaller and less consistent across topics (an increase of 0.16, 0.11, 0.09 SDs, respectively).

Aside from trust, we also measured people’s impressions of the expert’s specific attributes (e.g., How much time did they spend studying the topic? How much recognition do they receive from other experts? How relevant is their academic background?). People trusted an expert more to the extent they inferred that the expert spent a lot of time studying the topic, received recognition from their peers, and so on (Supplementary Table S1). Even though the attributes of experts were independently manipulated, people’s impressions of these attributes had small to moderate correlations among them. Furthermore, we observed a “spillover” effect, such that manipulating a single attribute impacted impressions of multiple attributes, in addition to the one that the manipulation targeted. For example, when an expert had an undergraduate major that was relevant to the topic, people also thought the expert had studied the topic longer, had graduated from a more prestigious institution, received more public recognition, and so on, regardless of the information they received about the latter attributes (Supplementary Table S2).

As with the first study, we observed context-specific effects for attributes such as gender (Fig. 2). People thought that a female expert had more personal experience on the topic of abortion and trusted her more. Replicating the importance placed on the expert’s wealth for stock investments, we found that for a stock expert, being a self-made multimillionaire was equivalent to having a degree in Finance (vs. Classics). Taken together, the first two studies revealed considerable overlap in what people said mattered and what actually mattered for their judgments, at least when they didn’t know the expert’s view on an issue.

Study 3

The first two studies showed that when people lacked information about an expert’s views, they trusted an expert with relevant academic backgrounds and more experience. We wanted to test whether or not people would still recognize those cues when the expert shared their strong, moralized opinions. We chose a moralized issue where people’s opinions split into two camps: abortion. Participants (n = 1,776; 940 pro-choice, 862 pro-life) read about a researcher who had written a best-selling book on abortion, who either had a medical degree and had spent a long time researching and publishing or one who had a mechanical engineering degree and had less research experience. In addition to the level of expertise, we manipulated the researcher’s views (i.e., pro-life vs. pro-choice vs. no information; we coded the first two as congruent vs. incongruent based on participants’ views) and gender. As pre-registered, we used the aggregated trust items (α = 0.96) as our main outcome and also measured impressions of specific attributes, similar to Study 2 (pre-registration for this experiment can be found here: https://osf.io/j6v4f/).

Overall, participants trusted the qualified researcher more than the less-qualified one (Fig. 3). They also trusted the researcher who shared their political views more than the one who did not. The effect of the researcher’s political views (η²_G = 0.195) was more than twice as large as the effect of expertise (η²_G = 0.075), and the two did not interact (Supplementary Table S3; see Supplementary Table S4 for results broken down for each trust item). The magnitude of the effect was striking given that participants in the first study didn’t rate an expert’s political preferences as highly relevant. Participants trusted a less-qualified researcher who shared their views (M = 5.05, SD = 1.36) just as much as they trusted a highly-qualified researcher whose views they didn’t know (M = 4.92, SD = 1.31). Similarly, participants trusted a highly-qualified researcher who held different views (M = 4.01, SD = 1.68) only as much as a less-qualified researcher whose views they didn’t know (M = 3.98, SD = 1.57). We found no evidence that pro-life and pro-choice participants differed in the strength of bias toward researchers who shared their views (Participant views x Researcher views interaction: F(2, 1759) = 1.72, p =.179), or in how well they recognized credentials (Participant views x Expertise interaction: F(1, 1759) = 0.93, p =.336; Supplementary Table S5). In line with our previous studies, people thought that a female researcher had more personal experience on abortion (F(1, 1763) = 4.73, p =.03). That did not translate into more trust in her across the board (F(1, 1759) = 3.45, p =.063). Instead, we found that only pro-choice participants trusted the female researcher slightly more, whereas we did not observe a gender preference among pro-life participants (Participant views x Gender interaction: F(1, 1759) = 12.23, p <.001; Supplementary Figure S1).

As expected, the effect of the researcher’s views on participant trust was larger among those who were more invested in their beliefs compared to those who weren’t as invested (Fig. 4). In other words, participants who had stronger moral convictions, confidence, and belief superiority made sharper distinctions between a researcher who shared their views compared to one who did not. Contrary to our expectations, being more invested in one’s beliefs did not interact with the effect of expertise when the researcher’s views were known (Supplementary Table S6). Participants distinguished between a qualified vs. unqualified researcher regardless of how much participants were invested in their beliefs.

Fig. 4 — Individual differences in confidence, moral conviction, and belief superiority moderating the effect of the researcher’s views on trust (Study 3).

We also replicated the “spillover” effect from Study 2 (Supplementary Table S7). Even though our expertise manipulation only entailed differences in years of experience and relevance of academic background, the manipulation also affected impressions of other aspects of expertise that were either not mentioned (e.g., personal experience) or fixed between conditions (e.g., prestige of academic institution) (ps < 0.001). Interestingly, we found that the researcher’s views also influenced all of the participants’ impressions (ps < 0.001). In other words, when people noticed that a researcher shared their views, they thought the researcher had spent more time studying the topic, had a more relevant academic background, and so on. We should note that the effect of the researcher’s views on specific impressions (0.014 < η²_G’s < 0.050; Supplementary Table S7) was much smaller than its effect on trust.

Finally, we asked participants to guess the researcher’s views in the condition where those views were not explicitly stated. Our prediction was that participants who were more invested in their beliefs would assume that the researcher shared their views, especially when the researcher was highly qualified. However, that was not the case (Supplementary Table S8). We also tested whether participants made guesses about the researcher’s views based on stereotypes. For example, participants might have assumed that the highly-qualified researcher would be pro-choice because educated people tend to be more liberal or that a male researcher would be pro-life. The researcher’s expertise or gender did not seem to influence people’s guesses, however (Supplementary Table S9). That suggests that people trusted good cues of expertise for their own sake, not because those cues allowed them to make assumptions about the researcher’s views.

Discussion

Our results reveal a complicated picture of laypeople’s evaluations of expertise. On the one hand, we found evidence that people are generally attentive to good cues of expertise (e.g., relevant training and experience) whereas they are less attentive to poor cues (e.g., ethnicity and height). We observed some variability in how much participants weighed those factors in their evaluations of experts across topics, but good cues were generally given more weight than poor ones, at least when the expert’s views were not known. That was true in both participants’ self-reports (Study 1) and their reactions to manipulations of various cues in a conjoint experiment (Study 2).

On the other hand, Study 3 revealed clear evidence of partisan biases in evaluating experts. When we specified an expert’s views on abortion, participants were significantly influenced by the expert’s opinions when determining the expert’s trustworthiness and credibility. While they remained attentive to good cues, the effect of congeniality with the expert’s expressed views was over twice as large as the effect of an expert’s having a relevant academic background; the discrepancy between the two effects was the greatest for perceptions of trustworthiness (Supplementary Table S4). In fact, participants trusted experts who agreed with them but had an irrelevant academic background (mechanical engineering) and limited experience significantly more than experts who disagreed with them but had a more relevant background (medicine) and substantially more experience. These results suggest a more pessimistic view of laypeople’s evaluations of experts on topics that are moralized and politicized.

The large effect size we observed for the expert’s views in Study 3 stands in contrast with the low self-reported relevance of experts’ politics in Study 1. It could be that people are unaware that their judgments can be affected by poor cues (including superficial characteristics like physical appearance) in the first place³⁸. Alternatively, people may consciously prefer likeminded experts, but be motivated to understate how much they value party allegiances over other good cues, in an effort to appear unbiased^39,40We did not explicitly ask participants to describe their thought process in Study 3, but it’s also possible that, when questioned, they might launder their partisan preferences through more acceptable cues they reported as relevant in Study 1, like moral character (e.g., “I trust the likeminded expert because they are less likely to deceive me.”). Some might acknowledge their partisan preferences more readily. Those who had stronger moral convictions, for example, not only showed more partisan bias (Study 3) but also appraised an experts’ politics as more relevant for expertise (Study 1) —in line with recent evidence showing that people are comfortable admitting to being biased for moral ends⁴¹. Future research should examine how people justify their preferences for likeminded experts, and whether there are lay beliefs about the nature of expertise that can be corrected by interventions⁴².

People’s impressions of specific cues of expertise were similarly complicated. We found that judgments about the relevance of an expert’s academic background or how much they are recognized by other experts largely tracked actual cues that we manipulated. But these judgments were also influenced by irrelevant cues. Some of this “spillover” can be explained by people making judgments beyond the information they have, based on reasonable assumptions. For example, when people see an expert who has published many articles and received many awards, they might assume that the expert has probably also received public recognition. But it is less reasonable to think that a mechanical engineer’s academic background should become more pertinent to abortion when the engineer agrees with one’s views, something which we observed in Study 3. It is likely that people initially formed a holistic impression of the expert and that impression then influenced more specific judgments—akin to the halo effect^43,44—thus allowing people to rationalize their preference for likeminded experts (e.g., “Why should I trust this expert who thinks like I do? Because they have a relevant academic background.”). Our findings show that while lay judgments of expertise might generally track normative standards of rationality, they can also be influenced by myside biases and coalitional motivations^20,21,45. While these biases can serve social functions, such as signaling group commitment, they can trap us in a closed cognitive loop, seeking experts who tell us what we already believe or want to believe. Future work should investigate how to decouple assessments of experts’ qualities from whether or not they agree with us.

We must acknowledge that our selection of stimulus materials might limit generalizability. We aimed to maximize the effect sizes when we manipulated credentials. But the differences between experts and non-experts might be less stark in real life. Although laypeople might find a mechanical engineer unqualified to offer medical advice, a chemist (as in the example of Linus Pauling) might be deemed “close enough” to weigh in. When we wanted to manipulate the expert’s political views, we picked a topic where most people already held opinions and observed a rather large effect. But for other topics—unlike abortion—that haven’t been the focus of intense policy debates and culture wars, expertise might prevail over politics.

Our results add important nuance to discussions of the “crisis of expertise.” While our participants were generally more attuned to good cues than the most pessimistic commenters might have predicted, we also found that our participants were prone to congeniality biases. These biases gave people inflated trust in ideologically “friendly” experts and deflated trust in ideologically “unfriendly” ones, independently of the experts’ actual credentials. Sometimes, people roughly know what qualities to look for in experts, but because they’re aware of how a suitable expert should look, they might invent or overemphasize a friendly expert’s qualities in order to rationalize their trust in that expert. Our findings thus reflect a social demand for congenial expertise, rather than a general, across-the-board distrust in expertise. In situations where no legitimate experts are willing or able to defend policy positions put forward by politicians (e.g., opposition to all vaccine mandates), laypeople who support those positions turn to “influencers” as sources of news and scientific information^30,46. Imagine for a moment that all of our current policy debates hinged on issues where there was genuine scientific disagreement and thus a lack of robust consensus. In that case, it’s easy to imagine there would be an adequate supply of ideologically friendly experts for all partisans. But even here the public would still be divided about facts, though without any “crisis of expertise” in some popular senses of the term. The current crisis, then, is not an outright dismissal of the value of expertise. Instead, it appears to be a crisis of selectively elevating or denigrating available experts in order to make one’s preferred views appear more justified. The old wisdom that “authority has a nose of wax” finds support in our data. And when everybody believes they are following authoritative experts, broad appeals to expert authority (e.g., “trust the science”) might fall flat.

Our findings do not reveal any readymade solutions for improving laypeople’s evaluations of expert cues in contexts of political disagreement and polarization. Indeed, we found that participant agreement with an expert’s opinion was twice as strong a predictor of participant trust than the expert having a relevant academic background. Unfortunately, that pattern of trust makes the public vulnerable to misbehaving or cynical experts. Our evidence suggests that epistemic trespassers²⁴might have social advantages over legitimate experts when they publicly express opinions demanded by an audience. While there might not be an epistemically ideal way to satisfy partisans’ demand for congenial expertise, our findings at least suggest that holding epistemic trespassers to account might help to improve partisans’ information supply. Due to the powerful social and economic incentives that some experts have for transforming into geniuses and oracles on the public stage⁴⁷, preventing them from trespassing is itself an ambitious goal. The best hope for an epistemic community might be to better legislate its boundaries and help direct laypeople toward the groups of experts who have the most legitimate claims to expertise.

Methods

All studies were carried out in accordance with the policies and procedures of the Human Research Protection Program (HRPP) as well as the ethical guidelines of the Human Research Protections unit of the Office of Research at the University of California - Irvine. Study protocols were considered exempt from review (Categories 2 and 3) by the Institutional Review Board at the University of California - Irvine (protocol #5756). All participants gave informed consent and received compensation.