Research and Ethics

Textbooks typically focus on the ethical treatment of participants when discussing ethical issues in research. While participant treatment is a central issue in research ethics, other ethical concerns -- often overlooked in textbooks -- are equally important to address. Let's assume research involves at least four distinct stages:

• Literature review
• Generating new hypotheses
• Gathering data
• Testing and reporting hypotheses

Each of these stages involves a number of ethical considerations. First, when conducting a literature review, two common ethical issues are plagiarism and failing to check primary citations. Plagiarism in any form is obviously unethical. Researchers should strive to ensure that they are not falsely misrepresenting others’ ideas as their own, and that they give appropriate credit for ideas and prior findings to their originators. Similarly, when generating new hypotheses, researchers should not “steal” others research ideas.

A second ethical consideration related to the literature review process is checking primary sources. Often, researchers will find a citation in one of the sources from which they are drawing information, and will simply cite that original work as the author cited it without reading the original source. This oversight has led to the false propagation of the idea that a cronbach’s alpha of 0.7 (a measure of scale reliability) is acceptable, "according to" Nunnally and Bernstein (1994). However, a careful reading of the original source reveals that Nunnally and Bernstein (1994) actually espoused that: 

“In the early stages of predictive or construct validation research, time and energy can be saved using instruments that have only modest reliability, e.g., .70. . . . In contrast to standards used to compare groups, a reliability of .80 may not be nearly high enough in making decisions about individuals. Group research is often concerned with the size of correlations and with mean differences among experimental treatments, for which a reliability of .80 is adequate. . . . If important decisions are made with respect to specific test scores, a reliability of .90 is the bare minimum, and a reliability of .95 should be considered the desirable standard” (pgs. 264-265). 

At no point do Nunnally and Bernstein assert that a reliability of 0.7 is universally acceptable. As you might expect, it is not uncommon for researchers to put a “spin” on others’ findings and theories in order to better support their own research and logic. Proper investigation of original sources is necessary for ethical, rigorous research and the correct attribution of conclusions.

Third, when gathering data, researchers must strive to protect participants from psychological and physical harm. As you all hopefully already know, the Institutional Review Board (IRB) was created to protect the rights of human research participants. Since the IRB is covered in depth in most research textbooks, we will not spend a great deal of time discussing its function here.

Finally, when testing and reporting hypotheses, a number of ethical considerations are salient, including honest reporting of decisions made during research (including elimination of “outliers”), the analyses conducted, the findings, and the limitations of the research. With respect to data cleaning, for example, a great many subjective decisions are made concerning subject selection and the removal of outliers. Any selective data analysis technique should be fully disclosed in research reports, especially as outliers are typically removed, and subjects selected for analysis, to maximize the chance of supporting the researcher’s theory. Further, when analyzing data, researchers are often taught not to “snoop” in their data (i.e., mining data for significant results) after having testing their hypothesis of interest, as snooping in data results in inflated Type I error. In other words, “significant” effects found during data snooping are more likely to be due to chance than would be expected given the set alpha level, especially without any alpha controls. However, since data collection involves a great deal of effort at the onset, some may argue that “snooping” is actually an ethical imperative --  that researchers should derive the most benefit from the efforts they invest in data collection. If you ever do yield to the temptation of data snooping, anything you find should be verified through replication of the research.

Other ethical issues arise in the reporting of research. One common ethical issue is the tendency to make misleading and inaccurate claims of causality, due to the importance ascribed to causal relationships in social science. Another ethical consideration in reporting is the extent to which researchers claim that their results are generalizable. Many times, samples represent a limited sub-population rather than the entire population of interest, especially given psychology's reliance on student samples. Researchers should be careful in generalizing beyond their sub-population, as the phenomena under investigation may be unique to a certain group of individuals.

Additionally, within the past several decades, a research technique known as meta-analysis has risen to prominence. Meta-analysis involves the quantitative synthesis and review of a body of literature on a given topic. According to some, meta-analysis is an ethical imperative for researchers as it makes the best use of existing data and may prevent unnecessary data from being collected. Rosenthal (1994) urges any researcher considering resolving a “controversy” in the literature to first conduct a meta-analysis to determine whether or not there truly is a controversy to resolve. However, while meta-analysis is a powerful technique and possibly an ethical imperative, it is not the panacea that some scholars assume it to be. Notably, meta-analysis involves a number of subjective decision points (e.g., definition of phenomena, what literature to include, including unpublished studies, measures accepted/collected, etc.). Consequently, two meta-analyses conducted on the same topic may result in different conclusions.

Finally, with respect to publication, Rosenthal (1994) highlights two additional considerations. The first is the general ethical obligation of researchers to share findings, either through publications, presentations, or citation of unpublished data. Rosenthal (1994) holds that censoring findings is unethical due to the time investment of the participants and researchers in obtaining data in the first place. The second ethical obligation is to ensure that authorship is assigned according to contribution, by some impartial ruling.

The above discussion is not intended to provide a comprehensive summary of all of the ethical issues involved in research – of which there are a great many. Instead, this discussion is intended only to provide an overview of some of the salient ethical issues throughout all stages of the research process. For additional reading on science and ethics, we recommend you read Rosenthal (1994).

References

Nunnally, J. C. & Bernstein, I. H. (1994).  Psychometric Theory (Third Edition).  NY: McGraw-Hill.

Rosenthal, R. (1994). Science and ethics in conducting, analyzing, and reporting psychological research. Psychological Science, 5, 127-134.