Web Usability Blog

Cognitive biases, gorillas and wine buyers

Written by Peter Collins | Nov 1, 2016 9:21:10 AM

Cognitive biases are the life blood of psychologists. These are ‘ways we think’ that mean we can make less than ‘rational’ decisions. Dozens of these have been identified from ‘Loss aversion’ to ‘Anchoring’ and I find it very satisfying when we find a good example of a bias. I noticed one recently when testing an on-line wine buying site. We identified that users did not understand how they could buy on this particular site i.e. a minimum of 6 bottles but then as many as they liked. Some felt they could buy less than 6, others that it was multiples of 6. The designers came up with lots of different ways to ensure the messaging was clear and prominent but still many users didn't get it.

What became apparent is that we were looking at a case of ‘Expectation bias’. Users were arriving at the site with a set of preconceptions based on their previous experience. “It is a grocery site so I can just add a couple of bottles” or “On-line suppliers always sell in cases of 6”. Users were not looking for messages about ‘how’ they could buy, they just got on with the task and bought some wine.

When we are task focussed we are very good at ignoring things we don’t think are relevant to the task in hand, as seen in the famous ‘invisible gorilla’ experiments of Simons and Chabris, where participants asked to keep score of a videoed basketball match did not see the gorilla in plain view in the video. This demonstrated ‘inattentional blindness’ as defined by Arien Mack and Irvin Rock in 1992.

For our task – buy some wine – users already ‘knew’ how they could buy (the expectation bias) so the messaging was just not seen: as the gorilla experiments show, it probably could have been bright orange and flashing and it would not have made any difference.

The solution was not to worry too much about ensuring users understood the proposition but to let then attempt to checkout and then deliver a suitable message – which they could not ignore because they had to dismiss it to continue and rectify the issue. What was interesting was that you might expect people to have been annoyed by this but they weren't – they were not unduly surprised to have their expectations, in these circumstances, challenged. So clearly the expectations were quite weakly held but, nevertheless, this still prevented them seeing something the proposition messaging – however prominent. So there is a clear message for site designers: if users are task focused and come with a set of expectations you can 'shout' at them as loudly as you like but it won't make any difference - if they need to get the message you will have to force them to read it with an accept/dismiss interaction.