Short answer: confirmation bias affects researchers by making them more likely to notice, value, interpret, or publish evidence that fits what they already expect, while overlooking evidence that challenges it.
Researchers are trained to be sceptical, but they are still human. They have theories, careers, expectations, funding pressures, and intellectual commitments. A scientist may genuinely want the truth and still be drawn toward evidence that supports their preferred explanation.
Where confirmation bias enters
Bias can enter when a research question is framed, when data are cleaned, when outliers are removed, when statistical models are chosen, and when results are interpreted. It can also enter through citation. Researchers may cite studies that support their view more readily than studies that complicate it.
Confirmation bias is especially dangerous when methods are flexible. If many analytical choices are possible, a researcher may unconsciously select the path that produces the expected answer. This is why transparency matters.
How science limits it
Science reduces confirmation bias through methods that make personal preference less powerful. Preregistration, blinding, control groups, data sharing, independent replication, peer criticism, and adversarial collaboration can all help. None of these tools makes researchers perfectly neutral, but they make bias easier to detect and correct.
The most realistic view is that science works not because individual scientists are bias-free, but because good scientific systems expose ideas to criticism. Confirmation bias is a reason to strengthen scientific method, not a reason to abandon it.