Prevalence of a Concept Changes How We View It

What's the science?

The term ‘creep’ describes the (sometimes unwanted) expansion of something over time, like the mission of a company or features of a product. It can also apply to abstract concepts such as aggression, a term which has expanded over time to apply to less aggressive acts than it used to. Many studies from neuroscience have taught us that humans judge things compared to their recent context (i.e. if there is less aggression, then an aggressive act may be seen as more aggressive than it did in the past). This can be a problem when the goal is to reduce the prevalence of something; if you are succeeding at getting rid of something bad, but keep calling a wider range of things bad, how will you ever know that you are making progress? This week in Science, Levari and colleagues investigate this phenomenon of ‘prevalence-induced concept change’ in humans.

How did they do it?

Participants were shown stimuli and were asked to determine whether a stimulus was an example of a particular concept. In the first experiment, participants were shown 1000 dots that varied from purple to blue. Over time, the prevalence of blue dots was reduced and participants’ responses were analyzed to see whether they called a wider range of colors blue (i.e. whether the concept expanded). A second condition with a stable prevalence of these concepts was also carried out as a control (i.e. the number of blue and purple dots was consistent over time). The authors then performed several replication experiments to test whether a) telling the participant that the prevalence would change, b) instructing them be consistent or c) changing the rate of reduction in prevalence would affect their behavior. They also tested what effect increasing the prevalence of a certain stimulus (i.e. increasing blue dots or purple dots) would have. They then repeated the experiments above with a more complex stimulus: faces that varied from non-threatening to very threatening. Finally, they performed another similar experiment in which they asked participants to judge the abstract concept of whether proposed scientific experiments were ethical or not. Participants were shown proposals for scientific studies and asked to determine whether they should be allowed.

What did they find?

Over time, participants were more likely to report a dot as being blue after the prevalence of the blue dots decreased. This result was robust, and persisted even after participants were told that the prevalence would decrease, and after they were instructed to be consistent. A more rapid decline in the prevalence of blue dots did not change this effect. Increasing the prevalence of the blue dots resulted in participants being less likely to report blue dots towards the end of the trials, demonstrating a reversal of this effect. When stimuli were changed to faces (a range of non-threatening to very threatening) the same effect was observed: participants reported more faces as threatening after a reduction in the prevalence of threatening faces. This also applied to the concept of judging whether a scientific proposal is ethical or not: the fewer unethical proposals presented, the more likely a participant was to reject a proposal as unethical. These behavioral results demonstrate a robust effect of “prevalence-induced concept change” that can apply to a variety of concepts.

concept1800.png

What's the impact?

This study demonstrates how widespread prevalence effects can be, ranging from simple color judgements to complex ethical judgments. Humans are likely to expand the definition of concepts when they become less prevalent. These findings have important implications for institutions that make decisions that need to stay consistent over time, like medicine or law enforcement. This phenomenon may help to explain why individuals fail to recognize progress, even as some problems really do get better.

conceptquote1800.png

Levari et al., Prevalence-induced concept change in human judgement. Science (2018). Access the original scientific publication here.