The Psychology of CAPTCHA

How Does The Brain See A CAPTCHA?

CAPTCHA’s are a popular way of stopping bots or machines interact with certain elements on a website. But how does a person’s brain react to a CAPTCHA? To understand this we need to consider how people make decisions.

In his book Thinking, fast and slow, Daniel Kahneman outlines how the human brain uses two different mental systems for making decisions:

System 1 – The fast, automatic, little or no effort, intuitive, but largely unconscious mind.
System 2 – The slow, disciplined, effort hungry, largely conscious mind that monitors system 1 and allocates attention to more complex mental problems that require it. However, this system is lazy and will rely on system 1 whenever it thinks it is adequately handling decision making.

As an example of how these systems work here are some simple puzzles. The answers are at the bottom of this page.

Do not try to calculate the answers, but listen to your intuition:

Question 1

A bat and ball cost $1.10 (one dollar and ten cents).

The bat costs $1 more than the ball.

How much does the ball cost?

Question 2

If it takes 5 machines 5 minutes to make 5 widgets, how long would it take 100 machines to make 100 widgets?

100 minutes OR 5 minutes

Question 3

In a lake, there is a patch of lily pads. Every day, the patch doubles in size.

If it takes 48 days for the patch to cover the entire lake. How long would it take for the patch to cover half of the lake?

24 days OR 47 days

These 3 questions made up the Cognitive Reflection Test that experimenters recruited students from Princeton to take. The questions were selected because they all suggest an immediate intuitive answer that is incorrect. For the bat and ball puzzle the number 10 (10 cents) tends to come to one’s mind. This is the intuitive answer but is wrong. If the ball did cost 10 cents and the bat is $1 more than 10c ($1.10). That would make the total cost of the bat and the ball $1.20, not $1.10 as stated in the puzzle.

Difficult to Read

When the test was administered, half the students were given a test paper with a small font and washed out print that was legible, but difficult to read. The other half were given a test paper in normal print. Interestingly, 90% of the students who received the test paper in normal font got at least one question wrong. However, only around a third of those given the difficult to read paper got any questions wrong. This is because the difficult to read paper caused cognitive strain (i.e. shit we have a problem!) which automatically activates system 2. This mobilises our full attention and allocates resource that is more likely to reject answers suggested by our intuition (system 1).

What this demonstrates is how easily we are happy to rely on our intuition (system 1), when things appear to be going well (i.e. no complex problems to solve). We rely on system 1 for making most our decisions, but this can sometimes cause us to jump to conclusions that are incorrect.

Implications for Conversion Optimisation:

If you want to ring alarm bells and activate your customer’s system 2 then make the font small and difficult to read. This may occur during the registration process for websites that use the Captcha security test (Completely Automated Public Turing test to tell Computers and Humans Apart). No wonder CAPTCHA may seriously damage your conversion rate! If you have been using persuasive copy to encourage sign-up, a CAPTCHA could undo all your hard work. It causes cognitive strain and activates a person’s system 2. This system is more likely to reject the idea of an impulsive decision and abandon a transaction if spooked by a CAPTCHA.

This model of decision making also suggests sites should avoid asking visitors to remember instructions or promotional codes etc (e.g. displaying codes as images that cannot be copied and pasted). The more information that a website expects visitors to remember for future use the more likely it will lead to cognitive strain. This will activate system 2. If mental effort is needed for storing information there will be less available for other activities and people become prone to missing messages or instructions during a website journey.

Focus Groups

The use of two different mental systems also challenges the way organisations use traditional research and usability testing for assisting website design. In a previous post, Why should you stop using focus groups?, I outlined why focus groups can be a misleading research tool.

However, standard usability testing, particularly in labs, are prone to some of the same kinds of bias. What Kahneman’s work suggests is that direct questions often engage the wrong system (system 2). Observation of behaviour is more likely to provide true insights. It also supports the benefits of ethnographic usability research where people are observed undertaking a behaviour in their natural environment (e.g. in their home) rather than in a user lab. Ultimately though the most reliable way of understanding what affects visitor behaviour will be an online experiment.