We make up to 10,000 decisions a day and although we may imagine that these are rational decisions based on our senses and the contents of our memory, probably most are made unconsciously using heuristics – strategies that use only a fraction of the available information. We just don’t have the time or brain processing power to do much else. Fortunately, most of the time the results are acceptable, however in certain situations they can lead to irrational behaviour and inaccurate judgment, especially when the decision involves uncertainty. These deviations from the norm are referred to as cognitive biases.
Since the early 1970’s, researchers have identified hundreds of cognitive biases, and although there is debate about whether some (or all) actually exist and the mechanisms and factors involved, there is no doubt that poor decisions are consistently made as a result of cognitive biases.
Reducing the negative impact of cognitive biases is a challenge due to the inherent nature of biases. Heuristics are believed to have evolutionary roots and have been deemed necessary for adaptation and survival. Our brains are wired in such a way as to make the operation of heuristics fast and automatic, creating in our minds a virtual “bias blind spot”. Work has generally focused on developing user training, typically scenario-based, to attempt to mitigate the effect of a small number of cognitive biases, but this approach has met with little measurable and lasting success. Other approaches (e.g. from intel, decision support and clinical domains) mostly involve adopting specific processes (e.g. using check lists) rather than computer-based tools. Research is now shifting toward modifying and improving the computer supported decision environment, especially within visual analytic tools, with efforts to ascertain the users cognitive behaviour. Other ongoing work is investigating the extent to which users are subject to cognitive biases when interpreting visualisations (in particular confirmation, anchoring and adjustment biases) and researchers have identified an attraction effect which can unduly influence their behaviour .
Some examples of cognitive biases which can be interpreted in the context of visualisation
Attraction effect (also known as the decoy effect and the asymmetric dominance effect) is where people tend to favour the option for which there exists a similar, but slightly inferior, alternative.
Anchoring is where the person is uncertain as to the answer and latches on to a value or narrative which is readily available in their memory, even if this is, in hindsight, irrational.
Adjustment is where the person is given an initial starting point (or result) and are reluctant to deviate too far from this value, thus inhibiting their exploration of the data.
Clustering illusion is where the user sees a pattern in the plotted data (e.g. on a scatterplot) when the data is in fact a random distribution.
Redundancy : Due to the fact that the visualisation can process and display a vast amount of data, the user perceives the result as being more accurate or correct than it actually is.
Availability is when people tend to use what is readily available (e.g what they can easily remember or what is on a display) and fails to search for further data
Sample : The smaller the data sample the less representative it becomes of the whole, but the display gives the impression that it is a better fit (e.g points all appear in a line).
 Dimara, E., Bezerianos, A., & Dragicevic, P. (2017). The attraction effect in information visualization. IEEE Transactions on Visualization and Computer Graphics, 23(1), 471-480.