A single exposure to cancer misinformation may not significantly affect related behavioural intentions

Introduction

While online information channels have the benefit of making health information more accessible^1,2, many are sceptical about the accuracy of online health information^3,4. Of particular concern is the prevalence of health misinformation online, this may be defined as health-related claims regarded as false due to a lack of scientific evidence⁵. Cancer misinformation is believed to be especially dangerous as it may encourage people to eschew evidence-based cancer treatments in favour of alternative cancer therapies^6,7, a decision associated with worse treatment outcomes. For example, a study which compared cancer patients who chose alternative therapies with patients who chose conventional treatments found that mortality rates among the alternative therapies group were 2.5 times higher after controlling for clinical and sociodemographic factors⁸.

Despite the recent proliferation of research into cancer misinformation and health misinformation in general, there has been little experimental work exploring possible causal links between health misinformation and health behaviour change. While many assume that misinformation exposure may lead people to believe the misinformation and act accordingly, decades of health psychology has shown that inciting meaningful change in health behaviour often requires going beyond information provision^9,10. For example, a recent review reported that attitude change interventions based on information provision typically exhibit small effect sizes¹¹. Even where these interventions are successful, attitude change is not always sufficient for behavioural change¹². Hence, although online health misinformation is held as a threat to public health, one should not take this for granted without evidence for an effect of health misinformation exposure on health attitudes and behaviours¹³.

Much of the existing research into the relationship between health misinformation and health behaviour has been conducted against the backdrop of the COVID-19 pandemic. Although numerous studies found that COVID-19 misinformation belief was associated with lower adherence to COVID-19 restrictions and reduced vaccination intentions^14–18, the cross-sectional nature of these studies makes it impossible to determine whether these relationships are causal. For example, it is possible that those who opposed COVID-19 restrictions were more likely to accept COVID-19 misinformation which trivialised the virus, and that this motivated reasoning accounted for negative associations between COVID-19 misinformation belief and support for COVID-19 restrictions^16,18.

Furthermore, controlled experiments which exposed participants to COVID-19 misinformation and measured its effect on behavioural intentions generated mixed results. Loomba et al. found that participants presented with misinformation about COVID-19 vaccines reported decreased intentions to get vaccinated against COVID-19 compared with participants presented with factual vaccine information¹⁹. Another study which exposed participants to various COVID-19 fake news headlines found that only certain headlines affected participants’ behavioural intentions¹³. For example, participants presented with a headline which suggested that a COVID-19 contact tracing app was recording their personal data expressed lower intentions to download the app. In contrast, a false headline which suggested that a COVID-19 vaccine was being released to the public despite high rates of side effects did not affect participants’ COVID-19 vaccination intentions. Thus, it is unclear to what extent health misinformation exposure can affect real-world behavioural intentions. To our knowledge, no experimental study has investigated the effect of cancer misinformation on behaviour or behavioural intentions.

Although it remains to be seen how much of an influence misinformation can exert on real-world behaviour, researchers have nonetheless sought to understand what makes people susceptible to misinformation and how the spread of online misinformation can be reduced. One of the dominant explanations is that people are motivated to believe misinformation which fits in with their prior beliefs^4,20. In recent years, however, a new cognitive-attentional account of misinformation has gained status, which proposes that susceptibility to online misinformation is not driven by motivated reasoning but rather, that it results from the tendency of online spaces to focus people’s attention on factors other than informational accuracy^21,22. Because online environments reward people for sharing novel, eye-catching headlines, people are thought to focus less on the accuracy of online information and more on its originality and intrigue. It follows that if people are primed to consider the accuracy of the information with which they are presented online, they will share less misinformation.

In a series of “accuracy-nudge” interventions, Pennycook and colleagues have demonstrated that when participants consider the accuracy of a single headline before reporting their sharing intentions for a series of true and false headlines, their sharing intentions for true headlines relative to false headlines are increased^22–24. These effects are present even when the initial headline has nothing to do with the later headlines. Despite this promising evidence, few independent replications of these results have been conducted²⁵ and it remains to be seen whether accuracy-nudge interventions only curb the spread of online misinformation or if they also dissuade people from engaging in behaviours encouraged by online misinformation.

These considerations led to the present study which measured the effect of exposure to various true and false cancer headlines on related behavioural intentions, as well as testing an accuracy-nudge intervention in a cancer misinformation context. Participants either received or did not receive an accuracy-nudge intervention before being presented with a series of true and false cancer headlines which mimicked the format of social media posts. Participants later reported intentions to engage in a selection of behaviours encouraged by the true and false cancer headlines. The study aimed to answer three research questions:

Methods

Materials

Accuracy-Nudge Intervention. The accuracy-nudge intervention was based on that of Pennycook et al.²⁴ Participants read three randomly-ordered real headlines unrelated to cancer which were circulated in the weeks before the survey was distributed - Vladimir Putin's rumoured authorisation of Donald Trump's 2016 election campaign, the UN's acknowledgement of a record high temperature for Antarctica, and details of Britney Spears’ conservatorship. Each headline was accompanied by a non-probative photograph (e.g., a picture of Britney Spears at an awards show). After reading each headline, participants answered the following question: "How accurate do you judge this headline to be?" Participants rated the accuracy of each headline on a scale from 0 to 100.

Cancer Headlines and Sharing Intentions. Each participant was presented with 4 true cancer headlines and 2 false cancer headlines in random order, with the false cancer headlines randomly selected from a pool of 4 headlines. The true headlines referred to known human carcinogens - HPV, processed meat, alcohol and sun exposure. The false headlines referred to unsubstantiated cancer treatments - cannabis and herbal medicines - as well as myths that fluoride increases cancer risk and that people with positive attitudes are more likely to survive cancer (see Figure 1).

Figure 1. False Cancer Headlines Used in Study.

After reading each headline, participants were asked the following question: "Imagine that you have a friend who has cancer. How likely would you be to share this headline with them?" Participants responded on a 5-point Likert scale ranging from "very unlikely" to "very likely".

Health Behaviour Intentions. At the end of the survey, participants reported their intentions to engage in eight health behaviours, each of which corresponded to one of the four true or four false cancer headlines (e.g., "I intend to smoke more cannabis"). Participants indicated their intention to engage in each behaviour on a 7- point Likert scale, ranging from "strongly disagree" to "strongly agree".

Results

All analyses reported were conducted at an alpha level of .05. No data were missing from analyses as only participants with complete outcome data were included in analyses. 387 participants were assigned to the accuracy-nudge intervention while 387 participants were assigned to the control condition.

As shown in Table 1, participants reported low to moderate intentions to share the false cancer headlines with a fictional friend suffering from cancer on a scale of 1 (very unlikely to share) to 5 (very likely to share). A paired samples t-test found no differences between mean sharing intentions for true (M = 2.29, SD = 1.08) and false (M = 2.31, SD = 1.09) cancer headlines, t(773) = -.571, p = .568, d = -.021.

To determine whether the accuracy nudge intervention affected participants’ sharing intentions for the cancer headlines, a mixed ANOVA was conducted on participants’ sharing intentions, with headline veracity (true or false) as the within-subjects factor and treatment condition (accuracy-nudge or control) as the between-subjects factor. There was no effect of either headline veracity, F(1, 772) = .33, p = .568, η²_p < .001, nor treatment condition, F(1, 772) = .45, p = .503, η²_p = .001, on participants’ intentions to share the cancer headlines with a fictional friend. There was also no significant interaction effect on participants’ sharing intentions, F(1, 772) = 3.79, p = .052, η²_p = .005.

As shown in Table 2, participants reported low to moderate intentions to engage in the four health behaviour intentions, on a scale of 1 (strongly disagree) to 7 (strongly agree).

Effects of misinformation exposure and accuracy-nudge intervention

To address our 3 key research questions, a series of four two-way ANOVAs was conducted, with fake news exposure (exposed or not exposed to the false cancer headline) and treatment condition (accuracy-nudge or control) as the independent variables and the related behavioural intention as the dependent variable. As shown in Figure 2, there was no effect of either misinformation exposure, F(1, 770) = 2.16, p = .142, η²_p = .003, nor treatment condition, F(1, 770) = .06, p = .804, η²_p < .001, on participants’ intentions to reduce their fluoridated water intake. There was also no interaction effect, F(1, 770) = 1.98, p = .160, η²_p = .003.

Figure 2.

Effect on Health Behaviour Intentions of (A) Exposure to Misinformation and (B) Accuracy-Nudge Intervention.

There was no effect of either misinformation exposure, F(1, 770) = .67, p = .413, η²_p = .001, nor treatment condition, F(1, 770) = 1.61, p = .204, η²_p = .002, on participants’ intentions to smoke more cannabis. There was also no interaction effect, F(1, 770) = .08, p = .779, η²_p < .001.

There was a main effect of misinformation exposure on participants’ intentions to adopt a more positive attitude, F(1, 770) = 13.23, p < .001, η²_p = .017. Participants exposed to the related false cancer headline (M = 5.17, SD = 1.37) were significantly less likely to intend to adopt a more positive attitude than participants unexposed to the headline (M = 5.53, SD = 1.22). However, there was no effect of treatment condition on participants’ intentions to adopt a more positive attitude, F(1, 770) = .18, p = .673, η²_p < .001, nor was there any interaction effect, F(1, 770) = .91, p = .341, η²_p = .001.

There was no effect of either misinformation exposure, F(1, 770) = .72, p = .398, η²_p = .001, nor treatment condition, F(1, 770) = .01, p = .932, η²_p < .001, on participants’ intentions to take more herbal medicines. There was also no interaction effect, F(1, 770) = 1.12, p = .290, η²_p = .001.

Discussion

The current study aimed to determine whether exposure to cancer misinformation could affect participants’ intentions to engage in related behaviours and to determine whether an accuracy-nudge intervention could reduce any such effect. Overall, exposure to cancer misinformation had a small effect on only one of the four health behaviour intentions – intentions to adopt a more positive attitude. The direction of this effect, however, was unexpected, as participants exposed to a false headline suggesting that people with more positive attitudes were likelier to survive cancer reported lower intentions to adopt a more positive attitude than participants who were not exposed to this headline. In addition, we found no effect of an accuracy-nudge intervention on participants’ intentions to engage in any behaviours endorsed by the cancer misinformation.

Although cancer misinformation and health misinformation in general are often cited as major threats to public health^4–7, this is the first study to investigate the effect of cancer misinformation exposure on related behavioural intentions. Therefore, it represents a significant step towards establishing the situations in which we should be most concerned about misinformation. Although curbing the spread of online health misinformation is itself an important undertaking, our results suggest that once-off exposures to cancer misinformation may not be enough to alter health behaviours. As established in the health psychology literature, the link between information provision and behavioural change is far from straightforward^9,11,12 and hence, one might expect that a relationship between misinformation exposure and behavioural change would be equally complex. When one considers that models of health behaviour include factors as varied as one’s self-efficacy, perceived susceptibility towards the related illness, and motivation to engage in the behaviour¹⁰, it becomes apparent that a single exposure to cancer misinformation might not significantly affect one’s intentions to engage in the related behaviour unless many conditions are met.

Furthermore, despite recent evidence in favour of accuracy-nudge interventions^22–25, we did not find any effect of our accuracy-nudge intervention on participants’ health behaviour intentions, nor on their intentions to share true or false cancer headlines with a fictional friend suffering from cancer. Hence, our results question the utility of the new cognitive-inattentional account of online misinformation susceptibility²¹. While others have reported that priming participants to consider accuracy reduces the amount of misinformation that they share²³, we found that participants were generally quite cautious to share the false cancer headlines with a friend suffering from cancer regardless of whether they received the intervention or not. In fact, no false cancer headline was “somewhat” or “very likely” to be shared by more than 50% of the participants exposed to it. However, there was still a substantial minority of participants who reported being “somewhat” or “very likely” to share the false cancer headlines with a friend suffering from cancer and hence, interventions which aim to reduce the spread of cancer misinformation remain important.

One of the main limitations of this study is that we did not measure health behaviours themselves but rather, intentions to engage in health behaviours. Although intentions represent a critical step towards behaviour change, decades of psychological research have shown that behavioural intentions do not always precede intended behaviours²⁸. However, since intentions are often the catalysts for behaviour change, we hold that since our false cancer headlines did not increase participants’ health behaviour intentions, it is unlikely that these headlines would have influenced behaviour.

Secondly, it is important to acknowledge that being presented with cancer misinformation in a study such as this is not the same as encountering cancer misinformation while scrolling through social media. While we emulated the format of social media headlines as best we could, participants likely use many different indices to judge the accuracy of a social media headline besides the content of the headline itself, such as the source who shared the headline and the number of people who “liked” the headline⁴. Removing this information may have changed the way in which participants judged the headlines, thereby making them less likely to affect their behavioural intentions. Furthermore, it is possible that real-world behavioural shifts in response to online misinformation only occur after multiple exposures to the misinformation. For instance, researchers have evidenced a reproducible “illusory truth effect”, whereby repeated exposure to the same information leads to it being perceived as progressively more accurate^29,30. Hence, it is possible if participants were exposed elsewhere to the cancer misinformation with which they were presented in this study, they would judge it to be more accurate and therefore be likelier to act in accordance with it.

In summary, we report that a single exposure to cancer misinformation is unlikely to change behaviour. While we support efforts to improve the accuracy of online cancer information and suppress online cancer misinformation, we note that certain claims about the potency of cancer misinformation may have been overstated. We therefore call for further experimental research to better understand the effects of cancer misinformation and the ways to reduce susceptibility to online misinformation.