Self-selection bias occurs when individuals select themselves into a group, causing a biased sample with non-random characteristics that can skew research results. It is a significant concern in observational studies and surveys, where the lack of randomization can lead to misleading conclusions about causal relationships.