While there is no firm consensus regarding the exact cause of CR, and undoubtedly different respondents have different reasons for their behaviour during surveys, most theories related to causes of CR can be grouped into either a lack of internal motivation or a lack of a sense of responsibility to respond carefully. As such, we sought to leverage approaches associated with well-supported social psychological theories to increase motivation of respondents and thus decrease CR. Results from three studies demonstrate that manipulating online survey instructions can reduce values on some metrics of CR. Study 1 found that video instructions designed based on social influence theory increased self-reported quality of participant data, scripted instructions decreased interest, and neither impacted objective indicators of CR. Study 2 found cognitive dissonance increased logical responses, interest, and response accuracy. Leveraging a sense of hypocrisy increased logical consistency and response accuracy, but not interest in the survey. Study 3 showed that leveraging social exchange theory in survey instructions generally had no effect on CR with the exception of reducing extremely variable responding (i.e. Mahalanobis distance values) when participants were offered a food/drink gift.
Taken together, findings from the three studies showed that logically consistent responding (Psychometric Synonyms), accuracy of responses (instructed-response sum), and interest (self-reported interest) were most influenced by instructional manipulations. Other types of CR were unaffected such as strings of identical responses (Maximum LongString), and time spent on the survey. The numbers of CR indicators affected by the design of survey instructions, suggests that manipulating elements of the survey shows promise as a way to prevent CR. Of course, there is much work to be done to understand CR's causes and prevent CR from occurring because several CR indicators showed no differences due to survey instructions. As with any exploratory study, replications of these findings are necessary to form stronger conclusions.
There are several possible explanations for the somewhat disappointing findings, with a few being particularly probable. First, there are numerous possible causes of CR (e.g. boredom or fatigue) and social psychological principles may simply not apply directly to CR behaviours in student populations. This reason may be particularly relevant to Study 3. We anticipated the importance of the salience of the experimental manipulations, conducted a survey of students from the same sampling population, and assessed what they would perceive as most valuable as a give-away that we could offer to them via survey. Although we pilot tested the experimental conditions that we were considering, to ensure that our manipulations would be salient to students, it seems that the impact of the manipulations was limited. It could be that our respondents simply did not care about a donation made in their name or a free item that may not have been desirable or convenient for them to obtain.
Second, the experimental manipulations we devised may have lacked psychological fidelity in some way. Although manipulation checks generally showed response patterns that we would expect from students in each respective experimental condition, there may be stronger manipulations with more impact on student behaviour. For example, reminding students of the instructional messages on each page of the survey would likely strengthen the effect. This would not necessarily need to be textual messages; rather, social influence could be induced through headshots or video messages from researchers and participants in other, similar studies. Alternatively, survey administrators could display to participants video testimonials from people whose lives were positively changed by results from research in which university students participated. This may increase the experienced task significance of students, and influence them to engage more highly in the task of completing the survey. This would follow directly from work design theory (e.g. Parker, 2014).
