What is confirmation bias and how strong is its effect?
Confirmation bias is the brain's shortcut to protect what you already believe. It's not just about ignoring evidence — it's about actively misreading it. In a 2022 experiment with nearly 3,900 people, researchers found that both pro-vaccination and anti-vaccination participants made systematic errors when evaluating data from a fictional vaccine trial that contradicted their prior beliefs [1]. The errors were statistically significant (P < 0.01), meaning they weren't random — they were driven by what people already thought. When the same people evaluated data unrelated to vaccines, those errors disappeared, confirming the bias was specific to their beliefs [1].
The effect shows up even in simple visual tasks. In a crowdsourced experiment, participants viewed scatterplots showing relationships between real-world variables (like environmental regulations and air quality). When they believed a strong relationship should exist, they overestimated the actual correlation by about 0.1 on an r-value scale (where 0 is no relationship and 1 is perfect). When they believed the variables should be unrelated, they underestimated by the same amount [4]. That's a meaningful distortion — roughly a 10% shift in how people perceive objective data, simply because of what they expected to see.
Why does this bias persist even when people know they're biased?
One surprising reason is that people often recognize their bias and approve of it. In a series of studies, researchers found that when people held a belief for moral reasons — like believing a friend is innocent — they acknowledged they lacked evidence but still felt their reasoning was ideal [5]. They even prescribed those biased beliefs to others [7]. This overturns the old idea that people are unaware of their biases. Instead, people sometimes think being biased is the right thing to do, especially when the belief feels morally important.
Another reason is that our brains approximate rational thinking in a way that naturally produces bias. A 2024 model called BIASR (Bayesian updating with an Independence Approximation and Source Reliability) shows that when people try to update beliefs rationally but have limited memory, they simplify by assuming beliefs are independent. This simplification automatically generates confirmation bias, including biased evaluation, attitude polarization, and belief perseverance [2]. So the bias isn't just a flaw — it's a side effect of how our cognitive system works under real-world constraints.
Can confirmation bias ever be useful?
Surprisingly, yes — in group settings, moderate confirmation bias can improve learning. A 2024 study using network models found that when individuals in a group had a moderate tendency to favor their existing beliefs, the group as a whole learned more accurately than if everyone was perfectly open-minded [3]. The reason is that moderate bias helps groups maintain productive diversity of opinion, preventing premature consensus on wrong ideas. However, the same study found that strong confirmation bias hurt the group's ability to find truth [3]. So there's a sweet spot: a little bias helps groups explore different angles, but too much locks them into error.
This doesn't mean confirmation bias is good for individuals. In political contexts, it weakens democratic accountability. A model of voter behavior showed that motivated reasoning has two harmful effects: it causes voters with different preferences to hold different beliefs (polarization), and it desensitizes voters to how well politicians actually perform [6]. The study found that desensitization — not polarization — is the main driver of reduced accountability. In other words, biased reasoning doesn't just make people disagree; it makes them less responsive to reality.
About These Sources
This answer is built on 7 peer-reviewed studies — published from 2021 to 2024, 2 from 2024 or later, 7 in Q1 journals, collectively cited 198 times — selected as the most relevant from 15 studies that passed quality screening, drawn from 61 papers retrieved from a database of over 500 million.
Sources used in this answer
Confirmation bias and vaccine-related beliefs in the time of COVID-19
In a 2x2 experiment with 3,899 participants, both pro- and anti-vaccine groups made systematic errors (P < 0.01) evaluating trial data that contradicted their prior beliefs, but not when evaluating unrelated data.
Confirmation bias emerges from an approximation to Bayesian reasoning
The BIASR model shows that confirmation bias (biased evaluation, assimilation, attitude polarization, belief perseverance) emerges naturally from a Bayesian approximation that assumes independence between beliefs to manage memory limits.
Can Confirmation Bias Improve Group Learning?
Network models show that moderate confirmation bias improves group learning, but strong confirmation bias harms the community's knowledge-producing capacity.
Seeing What You Believe or Believing What You See? Belief Biases Correlation Estimation
In crowdsourced experiments, participants overestimated correlations they believed in by about r=0.1 and underestimated correlations they doubted by about r=0.1 when viewing scatterplots with meaningful labels vs. generic X/Y axes.
People recognize and condone their own morally motivated reasoning
Across multiple studies, people recognized their own morally motivated biases and evaluated their reasoning as ideal, even when they thought they lacked evidence for the belief.
Motivated Reasoning and Democratic Accountability
A model of electoral accountability shows motivated reasoning harms accountability mainly through desensitization (weakening the link between performance and beliefs) rather than polarization.
Morality justifies motivated reasoning in the folk ethics of belief
Across three studies, people treat moral considerations as legitimate grounds for believing propositions unsupported by evidence, and sometimes prescribe evidentially poor beliefs to others.