When a person is trying to measure the association between an exposure and an outcome they might fall prey to this bias. It occurs when the exposure causes a third variable, let’s call it 𝒳. When the outcome also causes 𝒳, completely independently, this is when the bias may occur.
Seeing that the exposure and the outcome both cause 𝒳, and not realising that this is coincidence rather than a meaningful relationship, the observer might try to control for this collider, 𝒳. In doing so they will distort the actual relationship between the exposure and the outcome.

An example of this can bee seen in the ‘obesity paradox’. Obesity is the cause of cardiovascular disease (CVD) and of other health implications. These other health issues may themselves result in obesity, and also (independently of obesity) in CVD. They will also result in higher rates of mortality. When controlling for CVD it looks as though obesity is protective because there is a lower mortality rate for those with obesity than those with other ailments. You need only broaden the analysis beyond populations with pre-existing CVD to realise this is a distortion of the relationships.