Article Text
PDF
Abstract
This study experimentally tests whether the techniques of neutralisation as identified in the criminal justice literature influence graduate student willingness to engage in questionable research practices (QRPs). Our results indicate that US-born graduate students are more willing to add an undeserved coauthor if the person who requests it is a faculty member in the student's department as opposed to a fellow student. Students are most likely to add an undeserving author if a faculty member is also their advisor. In addition, four techniques of neutralisation, ‘diffusion of responsibility’, ‘defence of necessity’, ‘advantageous comparison’ and ‘euphemistic labelling’, are associated with student willingness to act unethically. Participants who had received responsible conduct of research training were no less likely to commit the violation than those who had not. Knowledge of these influencing factors for QRPs will provide for opportunities to improve research ethics education strategies and materials.
- Ethics
- Research Ethics
- Scientific Research
Statistics from Altmetric.com
Introduction
Responsible conduct of research (RCR) training generally assumes that knowledge of the principles of research ethics is sufficient to bring about ethical research behaviour. However, there are doubts about the usefulness of research ethics training for reducing the incidence of research misconduct.1–4 It is not clear if the increase in RCR training is leading to a corresponding reduction in incidences of research misconduct or ethically dubious practices among scientists. In fact, at least one study has pointed in the opposite direction—a recent study found that receiving training in research ethics led to no reduction in reported involvement with several types of what the study authors refer to as ‘problematic’ or ‘questionable’ behaviours (including improperly granting authorship credit), and was associated with an increase in some instances.5 To make research ethics training effective at preventing ethically compromising behaviour in the sciences, it would appear that merely requiring training in ethical principles and regulations is not likely to be sufficient.5 In addition, more research is needed into which factors lead to instances of these questionable behaviours and what can make these behaviours less likely. More specifically, our study tests whether the techniques of neutralisation, as introduced in the criminology literature, influence a person's willingness to add authors when authorship is not deserved. Authorship is an appropriate subject due to its centrality in the scientific credit/reward system and the prevalence of unethical behaviour.6 ,7
The situationist challenge to RCR
That ethics training is insufficient for improving behaviour would not be a surprise to those familiar with the situationist school of thought in social psychology. Situationism is roughly the view that knowledge about situations can be more informative than knowledge about character traits or dispositions for predicting behaviour.8–11 Seemingly trivial features of situations can affect whether or not someone is likely to cheat or to help others. For instance, in one famous study, people who found a small amount of money were significantly more likely to help a stranger.11 The challenge that situationism poses for RCR training is that knowledge and acceptance of current RCR principles will be insufficient to bring about responsible behaviour if situational features of the research environment heavily influence our ethical decision making.
A variety of goals, values and, possibly, character traits are at play in any given situation, some of which may at times conflict with each other. It may be that the person who improperly grants authorship credit is unaware that such actions are problematic or simply does not care. It is also possible that such a person does care about research integrity, but that other pressures and concerns (relationships with colleagues, debts to advisors and mentors, concerns about individual success) can swamp the strictly research ethics-related concerns or be enough to change the person's interpretation of the situation from being one involving questionable authorship practices to one involving collegiality.
Professional and ethical values and standards can come into conflict with each other. Even where there is no conflict, these principles often need further specification, leaving open the possibility that features of the situation, such as the actions of others, can shape people's interpretations of the situation and their choices of what to do. Therefore, to find an effective way of training scientists in the RCR, it is important to conduct further empirical research into whether and how these and other factors can affect behaviour in the sciences.
Factors influencing research misconduct and questionable research practices
For nearly 40 years, research ethics has been a public issue. Initial attempts to address the problem focused primarily on developing policies for identifying, reporting, investigating, adjudicating and penalising misconduct.12 This reactive approach gradually expanded to include the proactive approach of education in the RCR. More recently, researchers have been studying the causal, predictive and influencing factors for research misconduct, with the goal of informing both policy development and RCR instruction.13 ,14
A review of the literature shows a number of causal, predictive and influencing factors for research misconduct and questionable research practices (QRPs). These factors range from personality traits to situational features to institutional climate. Personality traits include egoism, narcissism, cynicism, a robust sense of self-entitlement, perfectionism, impatience and delusions of grandeur.15–17 Situational factors include competition among scientists for funding, conflicting roles in the workplace, and stress in personal or family life.16 ,18–20 Structural factors include features of institutional climate, such as the degree to which ‘publish or perish’ is reflected in an institution's productivity review process, and the degree to which institutions project a commitment to research integrity.21 These known factors reflect defects in personality, defects in transient circumstances, and defects in institutional climate; but there has been little research on defects in the decision-making process, namely avoidance of perceived threats of harm and processes of rationalisation.
In his retrospective study of closed case files, Davis reported data indicating that rationalisation plays a role in research misconduct, but Davis himself is unsure what his data actually show.16 ‘It seems that once offending researchers are caught, they tend to offer reasons for their behavior, many of which externalize the blame to others’.16, p. 411 Davis notes that these reasons are ‘similar to what Gresham Sykes and David Matza termed techniques of neutralization, which are formulated in advance of engaging in deviant behavior’.16, p. 411 However, Davis also notes that his study was unable to determine ‘whether respondents formulated these rationalizations before they engaged in research misconduct or only after they were caught’.16, p. 411 That is, because he was doing a retrospective study of closed case files, he was unable to determine whether the reason influenced the researcher's willingness to engage in misconduct or whether it was simply an excuse formulated after the fact. He also notes that his study was limited to case files of individuals for whom the Office of Research Integrity (ORI) in the United States Department of Health and Human Services has confirmed findings of misconduct. His study did not include cases of alleged but unproven misconduct, cases of misconduct that were never caught or cases that were confirmed at the institutional level but not reported to ORI because they were outside of ORI's jurisdiction.16
Recent studies indicate that the incidence of research misconduct and QRP is far more common than anyone would like to believe.14 Scientific misconduct includes fabrication of data, falsification of data and plagiarism (FFP). QRP include irresponsible behaviours, such as improper study design, altering or suppressing research results, misappropriation of resources, conflict of interest, careless peer review and irresponsible authorship practices.22 In addition to being unethical, FFP and QRP are undesirable because they compromise the integrity of science, increase costs, delay progress and decrease public trust. To prevent FFP and QRP, we first need to understand why researchers engage in such undesirable behaviour.
Our hypothesis is that, in many cases, researchers know the behaviour is wrong but find a way to justify it anyway. Certain thought processes and contextual features enable a researcher to neutralise an undesirable behaviour and interpret the behaviour as permissible or even desirable. We hypothesise that coercion and the standard techniques of neutralisation influence a researcher's willingness to engage in FFP and QRP by allowing them to engage in this undesirable behaviour without being troubled by their conscience.
Methods
To test our hypothesis, we use the factorial survey method to test the influence of coercive forces and Sykes and Matza's five techniques of neutralisation (defence of necessity, advantageous comparison, diffusion of responsibility, denial of injury and euphemistic labelling) on a researcher's willingness to engage in a QRP. Specifically, we suspect that graduate students are more likely to add an undeserving author if they believe that they have no other choice (defence of necessity); that what they are doing is not as bad as what others are doing (advantageous comparison); that doing so is necessary to compete with others (euphemistic labelling); that other graduate students are behaving the same way (diffusion of responsibility) or that their action will not hurt anybody (denial of injury).25
Participants
We collected data from 233 domestic graduate students from a large university in the Southeast USA. The study was approved by the university's Institutional Review Board. Each participant received 3 unique scenarios, yielding 699 possible responses. Mixed model analysis was used to test the hypothesis, as it is robust to the repeated measures issue of receiving multiple responses from the same participants. Of the 699 possible responses, 584 were usable responses from 201 participants. To encourage participation, we offered them the opportunity to participate in a drawing for an iPad Mini. If the students chose to participate, their contact information was collected in a separate data collection that could not be linked to the study data.
Approximately 40% of the responses were from male participants, while approximately 60% of the responses were from female participants. The participants averaged 30 years of age. Neither of these characteristics significantly (two-tailed p values) influenced the participants' decision to add an undeserving coauthor (gender: p=0.31; age: p=0.53). Twenty-five per cent of the participants have published articles, which did not significantly influence their decision to add an undeserving coauthor (p=0.55).
Experiment
The experiment was a scenario-based experiment of a factorial survey design. Instruments of a factorial survey design yield a crossed experimental design23 by using vignettes to present a realistic scenario with the tested factors randomly manipulated across multiple versions of the vignettes.24 Using vignettes, we provide the participants with a scenario in which the main character had just returned from presenting a paper at a conference and then faced the pressure of deciding whether or not to add an undeserving author to a paper. The first manipulation was the power distance of the undeserving author in the vignette: either the graduate student's faculty advisor, a faculty member who was not their advisor or a fellow graduate student. In the vignette, the undeserving author stated that they were responsible for the paper getting into the conference and should be added to the paper as compensation for accepting the paper into the conference. The second manipulation was the presentation of a technique of neutralisation statement that the main character was thinking as they were making the decision. The techniques of neutralisation were manipulated six ways: (1) no technique of neutralisation; (2) defence of necessity; (3) advantageous comparison; (4) euphemistic labelling; (5) denial of injury and (6) diffusion of responsibility.25 Each participant received three of the 18 vignettes (3×6 design) with the items from the two manipulations randomly inserted into the vignettes. See online supplementary appendix A for the vignette shell, sample vignette and the manipulations.
supplementary appendix
After reading each vignette, the participants were asked two questions. First, they were asked, ‘If you were the main character in this scenario, what is the likelihood that you would add that person as a coauthor?’ To reduce social desirability bias, the participants were also asked about the likelihood of another graduate student adding the undeserving coauthor. The responses were collected via a 101-point partially anchored scale ranging from 0=‘Definitely would not add them’ to 100=‘Definitely would add them’. The dependent variable was the average of two likelihood items.
Results
The model was developed using the PROC MIXED procedure in SAS V.9.4. PROC MIXED fits a linear mixed model and is robust to the repeated measures resulting from the participants receiving three vignettes, as it uses maximum likelihood estimates of variance to adjust for the possible correlation between the three observations from the same participants.26 The results (presented with one-tailed p values for directional hypothesis testing) indicated that US-born graduate students were more willing to add an undeserving coauthor if the person who requests being added as a coauthor was a faculty member (p<0.001) in the student's department as opposed to a fellow student (see figure 1 and tables 1 and 2). Students were most likely to add an undeserving coauthor if a faculty member was also their advisor (p<0.001). In addition, four techniques of neutralisation—‘defence of necessity’ (p=0.01), ‘advantageous comparison’ (p=0.04), ‘diffusion of responsibility’ (p<0.01) and ‘euphemistic labelling’ (p<0.001)—were associated with the students' willingness to act unethically. Denial of injury (p=0.10) was not significantly associated with the students' actions. Of particular interest to the researchers was that participants who had received RCR training were no less likely to engage in the QRP of adding an undeserving coauthor than those who had not received RCR training (p=0.46).
Participant demographics
Mixed linear model results
Likelihood of adding an undeserving coauthor and 95% CIs by techniques of neutralisation (grouped by power distance of the undeserving coauthor).
Limitations and discussion
Limitations
The study is subject to some possible limitations. The sample of 233 participants is approximately a 7.5% response rate. Possible non-response bias was tested by comparing the responses of early and late (those who did not participate until after a follow-up request) responders; no differences were found between these respondents. Therefore, the relatively low response rate is unlikely to bias the results.
Further limiting the study is that the responses were self-reported likelihoods of adding an undeserving author, thus leading to a possible social desirability bias. This possible bias was reduced by using, as the dependent variable, the average of the likelihood that the participants would add the undeserving coauthor and their belief that another person would add an undeserving coauthor.
Because this is a prospective study based on hypothetical scenarios, we cannot be certain that students in fact would add an undeserving coauthor if the specific situation described were to actually occur. However, as Davis16 notes, retrospective analyses of actual reported incidences of research misconduct cannot determine whether the rationalisations offered by respondents actually influenced the occurrence of the behaviour, or whether they are merely post hoc justifications. Looking at both prospective and retrospective studies can provide deeper insight into the causes of research misconduct and allow a greater degree of confidence in specific interpretations of the role rationalisations play in these behaviours.
Discussion
These results offer support for the hypothesis that techniques of neutralisation influence graduate student willingness to engage in QRP. All of the techniques except ‘denial of injury’ are associated with increased willingness to engage in irresponsible authorship assignment. The ‘denial of injury’ result is interesting because intuitively one reason why people might be willing to add an undeserving coauthor is the perception that adding another author does not harm anyone. Perhaps the sense of injury is enhanced in this vignette by adding another person to a single-authored paper. It would be interesting to see if students changed their perceptions about adding undeserving authors to papers already coauthored.
It is unsurprising that people were most likely to add an undeserving author when instructed to do so by their advisors. Students will use the techniques of neutralisation to rationalise their behaviour in situations in which they feel powerless to do otherwise. Advisors have a great deal of control over their students' educational and career prospects, leaving students in an unbalanced power relationship. The price of refusing an unethical request from one's advisor is high, which makes rationalisation more likely. One complicating factor is that advisors are often considered infallible sources of professional and ethical guidance, making some students believe ‘if my advisor tells me to do something, then that must be how it's done in our field and thereby ethical’. If true, some students might not be rationalising behaviour that they know is wrong, but changing their perception of the wrongness of adding an undeserving coauthor.
This explanation is consistent with some more recent interpretations of situationist research in psychology. Some have described situationist studies (such as Latané and Darley's9 and Isen and Levin's11 research on people's helping behaviours) as showing that objective features of situations are the predominant drivers of behaviour. Assuming that these studies do in fact reliably demonstrate an impact of ‘trivial’ features of situations on behaviours, the causal route is not necessarily a simple and direct one. How a person responds to a situation depends on the objective features of the situation itself, as well as on how the person interprets the situation. While proponents of strong situationism cite studies such as the Milgram obedience experiments27 and others as showing that robust, stable character traits (such as helpfulness or honesty) do not exist, critics of strong situationism argue that it is plausible that we look to cues from others to interpret what sort of situation in which we find ourselves. For example, we might look for clues from others to determine whether a situation is an instance where help is required. The actions of others can change our interpretation of the situation, and if it is the situation plus the interpretation of the situation that influences behaviour, then further research into the factors that affect these interpretations could also be useful. Our study supports this; the neutralisation techniques employed may be enough to settle someone on a particular interpretation of the situation in the presence of competing values and obligations. Since ‘euphemistic labelling’ was the neutralisation technique most associated with students' willingness to add an undeserving coauthor, it is possible that students are looking to other information beyond simply their knowledge of research ethics for clues about which actions are appropriate in the circumstances. When it is an advisor making the request, students may be more likely to interpret the situation as one in which their particular area of study has conventions and standards of appropriateness that do not fit neatly with ‘generic’ research ethics guidelines.
Other factors may also influence people's willingness to engage in QRP. For instance, Sabini and Silver28 argue that the Milgram experiments and similar widely cited studies indicate not that we overestimate the importance of character traits like compassion and honesty, but that we underestimate the impact our desire to avoid embarrassment has on our actions. Given a situation that is open to interpretation, the possibility that we might embarrass ourselves (eg, by overreacting or questioning a superior) can be enough to override impulses to do right or even make us more likely to settle on a morally neutral interpretation of the situation. In many social situations, our degree of confidence in our interpretation of what is going on and what response is required needs to be quite high to override these other concerns. Perhaps the current level and form of research ethics training is inadequate for instilling this confidence. This explanation would also be consistent with the findings of our study. Since students were more likely to add an undeserving coauthor when requested to do so by an advisor or another faculty member in the student's department, it may be that students are again either looking to more established members of the profession for guidance regarding standards that might be idiosyncratic to the particular field, or may lack sufficient confidence in their understanding of research ethics to be willing to oppose these requests. Anderson et al 18 found that mentorship was the only factor significantly associated with reducing the incidence of QRP. It may be that mentorship either helps shape people's interpretations of situations or helps instil the confidence necessary to overcome factors such as embarrassment. Further research on this would be useful.
It is interesting that participants who had received RCR training were no less likely to commit the ethical violation of adding an undeserving coauthor than those who had not. This finding suggests that RCR training (as currently taught) does not teach people how to resist pressures to engage in questionable behaviours. More research on the factors (situational and otherwise) that influence people's willingness to participate in QRP is necessary to discover ways to counteract these influences and design RCR training programmes that will be more effective at promoting ethical behaviour in the sciences.
Further research is also needed to determine: (1) the extent to which techniques of neutralisation affect other types of research misconduct and QRP; (2) whether presenting students and researchers with this information helps them identify the neutralisation processes in their own decision making and (3) whether, and in what way, RCR instructors can help students and researchers develop coping mechanisms to resist these influences toward undesirable behaviour.
This study focuses on subjects' willingness to add hypothetical undeserving coauthors; additional research could avoid this limitation by surveying the actual incidence of coerced authorship in research environments. Knowledge of reported cases of actual questionable practices will broaden our understanding of the underlying dynamics. Together, studies like this would provide a more detailed picture of the factors that influence peoples' willingness to engage in QRP. It would also be interesting to test whether subjects are more willing to add undeserving coauthors if they are promised reciprocity in return.
Broader knowledge of the factors that influence people to misbehave can improve research ethics education strategies and materials, allowing for research ethics training programmes to mitigate these challenges directly. For example, since ‘defence of necessity’ is associated with QRP, improved RCR training could explicitly highlight pathways to successful research showing how unethical behaviour can derail a scientific career. Similarly, since diffusion of responsibility is also associated with an increased willingness to engage in QRP, RCR educators could reinforce the relative rarity of bad behaviour in the sciences. Knowing that other researchers are not acting poorly makes it harder to justify one's own bad acts. On this line of reasoning, one way to promote good behaviour is to specifically design RCR educational materials that explicitly defuse the techniques of neutralisation that are allowing people to justify their questionable behaviour.
In addition, the data from this study will help instructors of RCR provide a more holistic approach to the topic of research misconduct. While the RCR curriculum typically defines FFP and QRP and provides estimates of their incidence, it is also important for the curriculum to include information about the causal, predictive or influencing factors for these behaviours. Failure to present this information perpetuates the ‘bad apple’ hypothesis, and encourages students and researchers to understand unethical behaviour as bad people behaving badly. More importantly, failure to discuss this information misleads students and researchers into believing that well-intentioned researchers simply need to identify and address FFP and QRP in other people's work. A more holistic approach to research misconduct will teach the causal, predictive and influencing factors for FFP and QRP, so that students and researchers can understand their own vulnerability to these undesirable behaviours. Presenting the standard techniques of neutralisation, along with data showing the ways in which these techniques can influence a researcher's willingness to engage in a QRP, will help students and researchers avoid unethical behaviour in their own work.
In summary, our study shows that four of the five techniques of neutralisation influence students' willingness to engage in the QRP of adding an undeserving author. Contextual features of the situations in which research is produced can have a significant impact on peoples' behaviour. Incorporating a discussion of the techniques of neutralisation into RCR training will help weaken their influence.
References
Footnotes
Contributors BST, TP and BM conceived the study. BST designed and ran the study, and analysed the data. BST, TP, AH and BM interpreted the data, wrote and edited the paper.
Competing interests None declared.
Ethics approval Mississippi State University Institutional Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.