Confirmation bias is the tendency to favour evidence that supports existing perceptions.
Confirmation bias is often discussed in the context of science. The problem is even worse in finance because it is not possible to replicate can freely vary experiments to find new evidence — we have to rely on available real-world evidence. This means that the best defence against confirmation bias lies in proper back-testing, and simply taking care.
The complexity of many financial models also tends to create opportunities for bias to creep in. Although evidence might not actualy be over-looked completely, there is a tendency to tweak models towards expected results. For example, a sell-side analyst whose forecast differs from the market consensus is far more likely to review their assumptions than one whose forecast is close to the consensus. The consensus thus reinforces itself. Even worse, is that know the consensus while making a forecast will skew every judgement made while developing a model.
It does not help that the rewards of being wrong on call like that far out-weigh the rewards.
One defence is to simply avoid looking at at others' results while working on a model. The best is simply to cultivate the habit on independent thinking, which is hard.