A Contingency Learning Approach to False Belief Formation

Student thesis: Doctoral ThesisDoctor of Philosophy


The prevalence of fake news in the current era, is thought to lead to the formation of false beliefs (Ecker et al., 2022). The false beliefs that arise from fake news are dangerous as they often guide decision around health, politics, and finance (European Commission, 2010). However, attempts to remedy false beliefs by exposing people to facts have largely failed to work (Yarritu & Matute, 2015). As such, a systematic, fundamental understanding of the mechanism through which false belief formation occurs is essential to develop effective countermeasures against fake news. Contingency learning provides one framework to understand the development and maintenance of some false beliefs. Contingency learning paradigms can be utilised to further understand the formation of associations between events when there is no true relationship between the events. In this thesis, I investigate these contingency learning principles and put them in context of false belief formation, especially, but not exclusively, in relation to medicine and disease.

Currently, the majority of contingency learning studies utilize active paradigms. These active paradigms require people to be actively involved in the process of responding to individual stimuli before their perceived control over the outcome is measured. Less attention has been paid to passive contingency paradigms. In these, people instead of directly interacting with the stimuli, simply witness the stimuli and outcomes appear, followed by a measure of their perceived relationship. Despite being less prevalent in research, these passive contingency paradigms are possibly especially interesting to study in context of false belief formation in society, which, for many, may be the results of associations formed through observations (e.g., being exposed to social media information) rather than actively interacting with this information. I adopted this passive contingency paradigm accordingly.

In Chapter 1, I outline and discuss a range of false beliefs, and then narrow down to false beliefs formed from causal illusions. Chapter 2 discusses the methodological framework of the contingency learning literature in light of the methodology of the seven experiments used in this thesis. This review then led to the design of the seven experiments that forms the empirical backbone of this thesis. The first 5 experiments considered the boundary conditions under which density effects occur in context of judgments of effectiveness of putative treatments for disease. These five studies thus helped to uncover what may prevent or exacerbate false beliefs, and simultaneously addressed methodological expansions of established contingency learning paradigms. These initial five studies focused on beliefs that were newly formed; there was no a priori knowledge about cue and outcome prior to the experiments. The final two experiments, in turn, examined the outcome density effect with regards to the broader context in which they may form: the role of pre-existing dispositional affect, and the existence of prior beliefs. These seven experiments are reported in the chapters that follow.

In Chapter 3 I investigate the influence of outcome density on judgements of effectiveness of medication, in a contingency paradigm, to test if outcome density effects differ between online and lab settings (Experiment 1). Large effects of outcome density were found across both settings, suggesting that participants give higher judgement of causality when the probability of outcome is higher compared to when the probability of the outcome is lower. In Chapter 4, I examined density effects by comparing cue and outcome density effects directly, to see if the same cognitive principles underlie both (Experiment 2). Effects of both cue and outcome density were found, however outcome density effects were larger than cue density effects.

To further understand these effects, in Chapter 5, Experiments 3 and 4 were conducted in which the order of cue and outcome presentation were manipulated, along with cue and outcome density effects. This extension further compares the density effects and allows for the exploration of whether the effect of cue or outcome density on contingency judgment is altered by reversing the order in which event information is seen. The size of the density effects stayed relatively consistent across the presentation order manipulation, and outcome density effects remained larger than cue density effects. The order of event presentation seems to make little or no difference to outcome and cue density effects.

In Chapter 6 I further examined the outcome density effect across positive, negative and zero contingency conditions within the same study to compare the effect of outcome density (Experiment 5). In addition to this, the effect of scale length was explored by comparing a frequently used unidirectional judgement scale to a bidirectional scale. I found effects of outcome density, contingency, and scale. In addition to this, people naturally seem to be making associations that are more positive than the contingency would indicate.

In chapter 7 and 8 I examine the role of prior beliefs (Experiment 6). The influence of pre-existing prior beliefs on judgements in passive contingency tasks has not been explored thus far, as far as I am aware. In addition to this, I examined whether the outcome density effect occurs in different contexts. Chapter 7 covers a pilot study, and Chapter 8 then builds on that through an experiment. There was some variability in judgement and prior beliefs across contexts. Overall, there was evidence to suggest that context-specific prior beliefs influence contingency judgements.

In Chapter 9, the final empirical chapter of this thesis, the role of affect is examined (Experiment 7). The contingency literature has focused on the depressive realism effect within active contingency paradigms. As such this experiment extended the research by examining a wider range of affect to identify whether naturally occurring differences in emotional state relate to differences in judgment. Affect was found to influence contingency judgements. Specifically, in the low outcome density conditions there was an influence of positive affect, negative affect, and anxiety on contingency judgements. In the high outcome density conditions, there was an influence of anger and stress on contingency

Finally in Chapter 10 I discuss the findings in terms of their theoretical, empirical, and methodological contributions. In addition to this, I also provide recommendations for future areas of research within this field and discuss strengths and limitations of the findings.

Together, the work presented in this thesis has enhanced the understanding of the mechanisms under which false beliefs may be formed from causal illusions and provides potential implications for current interventions on reducing false beliefs as well as future
Date of Award1 Nov 2023
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorTim Rakow (Supervisor), Nicola Byrom (Supervisor) & Wijnand van Tilburg (Supervisor)

Cite this