The Morality of Statistics: Ethics, Lies, and Responsibility

Introduction: Can Statistics Lie?

A famous book from the 1950s was entitled How to Lie with Statistics. Implied by its counter-intuitive title was the idea that the old cliché — "numbers don't lie" — is false. In fact, as discussed in the article "Reflection before Action: The Statistical Consultant Confronts Ethical Issues" by S. Andrew Ostapski and Claude R. Superville, statistics can be highly subjective in terms of how they are presented, as are the conclusions that can be drawn from them. Even researchers have been accused of manipulating statistics to prove "facts" that are not true within academia. The pressures only increase when statisticians are asked to serve the financial interests of commerce. "The ability to be creative in building interdisciplinary bridges can be risky, especially when the parties that are served do not understand the statistical process. The statistician must not only make sense out of the data but also develop the means to ensure the proper interpretation of such information by all relevant parties" (Ostapski & Superville, 2001).

For example, a statistician may be solicited by a company wishing to prove the effectiveness of a new workout program — demonstrating that statistically significant results occur when persons use an exercise machine developed by the company. However, there may be many problems with the designed study. The experimental group may have been placed on a strict diet while the control group was not. Moreover, the fact that the group using the workout machine experienced better results than the control group might be attributed simply to the fact that some form of exercise is better than none — not that people need to purchase the expensive device to achieve the same results. Finally, experimental research must also account for the observer effect: the mere act of being observed may cause participants to change their behaviors. Without acknowledging these factors, claiming that the program was effective does not paint a full picture of the results.

Statisticians do not have a formal code of professional ethics comparable to those in law or medicine, so there is a degree of subjectivity in the decisions they make. Yet the impact of statistics can have far-reaching consequences if deployed unethically. A study that obscures the potential side effects and dangers of a drug could cost millions of lives; public policy decisions, too, are regularly made on the basis of statistical data.

Ethical Frameworks for Statistical Practice

A Kantian or deontological approach to ethics suggests that a person must ask whether there is a duty to act, and whether they would be willing to act in the chosen manner as if setting a universal law for all time. In contrast, a utilitarian approach applies a cost-benefit analysis — do the benefits of a particular statistical approach outweigh the harms? Still other methods of assessing ethical behavior include community standards (would "the decision be the same if it were published on the front page of The Wall Street Journal?") and the Golden Rule (would such actions be acceptable if "done unto me?") (Ostapski & Superville, 2001). The varied application of these different ethical rubrics is likely to produce very different conclusions about what constitutes moral statistical practice.

Deontological vs. Utilitarian Approaches to Data Reporting

If one takes a deontological view that all misrepresentation is wrong, then any statistical analysis that does not paint a full and forthright picture of the matter being studied is immoral. The much-touted statistic that "9 out of 10 dentists recommend" a particular brand of toothpaste would be ethically questionable under this view, even though the consequences are arguably minor and constitute accepted advertising "puffery." Even statistics that are technically correct but likely to be misread could be considered unethical if they motivate people to act in questionable ways.

Consider vaccine side-effect reporting during flu season. It is widely regarded as beneficial to public health for as many eligible people as possible to be vaccinated. However, there will always be some reported side effects that may or may not be directly attributable to the vaccine itself. A deontological position holds that the statistician has a responsibility to report the truth — even if doing so may discourage some people from being vaccinated — because there is a well-documented psychological tendency to over-value the negative effects of action compared to inaction. A Golden Rule analysis might similarly support full disclosure, since most people want to know about potentially adverse consequences, however slight.

In contrast, a utilitarian view might suggest that the potentially severe consequences of widespread vaccine refusal are so great that side-effect statistics should not be prominently reported — or should be presented in a dismissive manner in fine print — to avoid undue alarm. However, the same utilitarian would likely struggle to justify suppressing statistics about the serious side effects of a profitable drug when those side effects are real and the drug offers no meaningful benefit over safer alternatives. In such a case, reporting suppression serves profit rather than public welfare, and the utilitarian calculus shifts decisively against it.

Conclusion: Statistics, Free Will, and Personal Ethics

On a personal level, the analyses of both Ostapski and Superville and Geertsema serve as a reminder of the limits of objectivity, even in a numerically driven science. An understanding of both the principles behind the analytical method and the consequences of its interpretation must be taken into consideration. The Christian belief in human rationality must also be tempered by an awareness that humans are fallible and prone to misunderstanding — particularly when fearful or self-interested, as is often the case when people seek information from business-related sources. Ideally, statistical evidence should support people's ability to make free and informed choices, not undermine it.

References

Geertsema, J. (1987). A Christian view of the foundations of statistics. Perspectives on Science and Christian Faith, 39(3), 158–164.

Ostapski, A., & Superville, C. (2001). Reflection before action: The statistical consultant confronts ethical issues. Business Quest. Retrieved from http://www.westga.edu/~bquest/2001/consultant.htm