Overcoming the Era of #FakeNews: A Contingency Learning Approach

Lead Research Organisation: King's College London
Department Name: Psychosis Studies

Abstract

In the age of the Internet, the increased prevalence of fake news is supporting an increase in beliefs that are not founded on evidence (Carroll, 2015). Algorithms used by social media and search engines exacerbate the impact of false beliefs by favouring information that echoes people's cherished ideas and beliefs. The resulting myths become increasingly difficult to eradicate (Lewandowsky et al., 2012). Research indicates that showing people facts is not enough to eradicate false beliefs (Yarritu & Matute, 2015). Clearly, a systematic, fundamental understanding of belief in fake news is essential to develop effective counter-measures.
At the root of false beliefs, such as those propagated by fake news, lies the phenomenon of contingency learning: the process by which people learn to associate one variable with another. My objective is to understand this process in context of false beliefs. By doing so, I will develop a strategy to prevent false beliefs from forming and, consequently, prevent people from sharing these fake stories with others.
Contingency theory (Allan, 1980) provides a normative account of causal belief, stating that the belief should correspond with the difference in the probability that two events co-occur, P (Event 1|Event 2), and the probability that one of these events occurs in isolation, P (Event 1|No Event 2). This difference is quantified as the degree of contingency.
Unfortunately, people do not adhere to this normative account; contingency beliefs are often biased. People tend to believe that events that occur often are somehow related, even if they are not (Allan & Jenkins, 1983; Byrom et al., 2015). People also use irrelevant contextual cues to infer causality (Van Tilburg & Igou, 2014) yet fail to consider information relevant to their judgement (Matute et al., 2015). These discoveries in contingency learning research offer an ideal basis for understanding why people believe fake news, and what can be done to counter these beliefs.
Many people believe that taking Vitamin C cures the common cold. A systematic review found that there was no trial evidence for the efficacy of Vitamin C (Hemila & Chalker, 2013). As people tend to recover with or without Vitamin C supplement, they are at risk of incorrectly attributing their recovery to Vitamin C.
The relationship between Vitamin C and the common cold is a low-cost scenario. However, the situation becomes troubling when considering the unconfirmed belief that cannabis oil is a natural treatment for cancer. Facebook pages echoing this claim are followed by many thousands. People have inferred a correlation between cancer remission with taking cannabis oil and may choose to use cannabis oil as a treatment for cancer, which have had disastrous health consequences.
Research into false beliefs by examining their roots in contingency learning may provide crucial insight into understanding and preventing the negative impact of fake news. To date, most work has been focused on either scenarios where participants make judgements of their control over events or they passively observed the relationship between two events.
I will test if and how contingency learning shapes belief in fake news, by examining how people evaluate the truthfulness of a 'news headline' that claims a statistical relationship between two events (e.g. "cannabis oil cures cancer") on basis of information presented to them. For example, participants will be presented with cases of hypothetical individuals who are described as being administered cannabis oil or not, and as experiencing a remission in cancer or not. Similar headlines on different topics will be used to assess the universality of effects. This approach places the contingency learning paradigm in a context directly relevant to understanding belief in fake news. Based on these results a second pair of experiments will test how well and long effect of training to identify fake news will last.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
ES/P000703/1 01/10/2017 30/09/2027
2104630 Studentship ES/P000703/1 01/10/2018 30/10/2022 Sahana Shankar