Skip to content

Oxford Uehiro Prize in Practical Ethics: Can Science Ethically Make Use Of Data Which Was Gathered By Unethical Means?

  • by

This essay was the runner up in the undergraduate category of the 6th Annual Oxford Uehiro Prize in Practical Ethics

Written by University of Oxford student Toby Lowther

In this paper, I discuss the question of whether science can ethically make use of data which has been gathered by unethical means in seeking scientific and medical advances to alleviate future suffering. This is an ever-controversial issue of practical ethics, and although the American Medical Assosciation provides firm guidelines on the matter (AMA, 1995), the ethical question remains complex. I will begin by laying out the core issue: the conflict between the desire to censure unethical practices used in gathering such data and the desire to use all data available to bring about the greatest good for society. I will present arguments either side, leading to an ethical stalemate, before presenting how issues of practical consideration for scientific methodology resolve the conflict. I conclude that science cannot make use of data gathered by unethical means, because such data cannot ethically be replicated, and reproducibility is necessary for the validity of the scientific method. I leave open the question of whether it is ethical for the findings of such unethical experiments to guide future, ethical research.The root of the problem is this: advances in science often bring about great good, such nuclear power, radiology, and so on. Scientific advances are more frequent and more reliable when science draws upon a broad basis of good data. Hence, it appears there is an ethical imperative for scientists to make use of all data that are available to them to drive scientific advance. However, what are we to say when this data itself was gathered by unethical means?

By data that is gathered by unethical means, I mean data which could no longer be morally gathered under the present understanding of ethics in science. Canonical examples would be the data from Milgram’s shock experiments (Milgram, 1963), the Stanford Prison Experiment (Haney, Banks and Zimbardo, 1973), much of Harlow’s work (e.g. Harlow, 1958; Harlow, Dodsworth and Harlow, 1965), and in the most extreme case, the ‘biological’ experiments of the Nazis on the victims of concentration camps. That these experiments were unethical is undeniable: Milgram lied to his participants, denied participant withdrawal, and left some participants severely traumatised; the Stanford Prison Experiment (henceforth SPE) had to be shut down it became so dangerous; Harlow was engaged in rampant and systematic animal abuse; and the extreme immorality of the Nazi experiments I take as evident without comment. Yet some of these experiments also produced usable data: Milgram’s study is still influential in studies of obedience in social psychology; Harlow’s monkeys demonstrated apparently effective models of isolation and depression, and of human behaviour in returning to abusive relationships; and the data from Nazi experiments have been used in many branches of modern medicine. All of this understanding could potentially help to save lives and improve society, if taken through science into practice.

This then is the dilemma. Upon the one hand, if we use the data, we appear to be condoning the unethical and immoral practices that allowed the data to be gathered and produced. Upon the other, if we do not use the data, we risk delaying scientific and medical advancement, potentially causing unnecessary further suffering.

The ethical argument against use of such data can be summarised as follows. To allow data which is gathered by unethical means in scientific investigation validates gathering data by such means. Firstly, this appears to justify the original unethical actions of the researchers. We want to censure those who engage in unethical research practices, but it is difficult to do so if these unethical practices produce data that is used in advances that alleviate suffering, as the unethical research practices appear as direct causes of the alleviation of suffering. Secondly, it risks emboldening researchers to engage in such unethical practices, by affirming the value of data gathered by such practices and by giving the guarantee that such data will not be ignored, even if the researcher themselves may come under censure. Thus, the argument goes, we should not use such data in censure of unethical research practices.

However, there is a strong argument against this position: namely, that ignoring the data does not undo the crimes by which it was gathered, but it may result in further harm by delaying or preventing scientific and medical advances which could alleviate suffering in the world. The argument here is not a sunk cost fallacy: it is not to claim that ‘we’ve got this far’, so may as well keep going with the data we have. Rather, it would be unethical for a scientist to ignore data which, if considered, could lead to or hasten developments in science and medicine that would alleviate suffering. The root of this argument hangs upon the consequentialist intuition that it is the effect, not the means by which it is reached, which is important for the ethics of an action. If a scientific advance leads to great benefit to humanity (in technology, medicine, psychology, and so on), and does not itself cause harm, then the mere fact that the data used to reach that scientific advance was gathered by unethical means does not prevent pursuit of that scientific advance being the ethically correct action.

If we assume a consequentialist ethical framework, it appears that we are left with an issue of epistemic access. It appears that if using the data would lead to greater harm to the degree that it reinforces unethical practices, we should not use it; and if it would lead to greater good to the degree that it enables or hastens advances in science and medicine, we should use it. However, we cannot know in advance whether making use of such data would necessarily reinforce unethical practices, nor can we know in advance whether use of such data would necessarily enable or hasten advances in science and medicine that would bring about good. The ethical argument appears to reach a stalemate, for it depends entirely upon the prior probability we assign to individual outcomes of the use of such data.

Fortunately, in matters of practical ethics, we are not limited to sole consideration of the abstract ethical values of acts, but we can supplement such considerations by practical considerations, and here, it seems, we can find a resolution to our present issue. For science has a methodology, which grants it rigour and by which it gains its epistemic value. If we can establish that such unethical data cannot be ethically accepted within the values of the scientific method, our stalemate would be resolved.

A first proposal may be to look to methods employed and ecological validity. The SEP has been widely criticised for the direct involvement of the researcher in the experiment, which is likely to induce bias in the results (Gray, 2013). Similarly, issues of ecological validity have been raised in moving from Milgram’s laboratory to society as a whole, and from Harlow’s rhesus monkeys to human neuroanatomy and psychology. These are certainly valid and important criticisms, and concerns of soundness of methodology and ecological validity should be raised against any data which is to be used in scientific progress.

However, I do not believe that these criticisms by themselves can guarantee a universal answer to our question. For it is conceivable that some of the data may be have been gathered with perfect methodology, so far as every question other than ethics is concerned. Indeed, my own reading of it would suggest that Milgram’s study is as sound as any social psychological study of his time, and does a fair job of explaining, in behavioural terms, a human tendency that has allowed the many genocides of the last century to come about. But let us set aside Milgram and suppose an idealised case, in which we have a study which was: methodologically rigorous, ecologically valid, producing data relevant for development in science and medicine which could potentially reduce suffering greatly, but which was unethical in practices. What are we to say of this supposed study?

Even in this case, I believe we have reason to reject the data from this study upon the basis of one central principle of science: reproducibility. The demand of reproducibility in science has come under intense scrutiny recently, with the growing ‘Reproducibility Crisis’ in the social, cognitive, and biomedical sciences. The principle behind espousing reproducibility as a central virtue of science is the claim that science allows us to reliably produce knowledge precisely because the findings of science are replicable, and replication allows us to confirm our theories with greater surety. Whilst a number of researchers have questioned universal demands for reproducibility (e.g. Drummond, 2017), it is not at all clear what alternative may be offered to guarantee that a particular effect does not stem from a sampling error in the study at hand. Without reproduction, we can never confirm whether the effect that appeared in the data was generalisable to the population or merely an idiosyncrasy of the particular sample under consideration. Hence, replication and reproducibility form the founding cornerstone upon which the validity of science is built.

It is clear how this relates to the question at hand. Data which has been gathered by unethical means cannot ethically be replicated: to do so would require repeating the unethical experiment. Hence, it is not ethically reproducible. As it is not reproducible, it is therefore inadmissible as evidence into the empirical developments of science, which rely, fundamentally, upon the principle of replication.

Where does this leave our question: can science ethically make use of data gathered by unethical means? The answer we have reached is a complex ‘no’. Science cannot make use of data gathered by unethical
means, because it cannot ethically replicate such data, and hence, the data violates the principle of replication which lies at the heart of the methodological rigour of science. This conclusion resolves the stalemate we reached above: for it is no longer relevant whether it would be in principle ethical to make use of such data, for we cannot, in practice, without violating one of the central constraints placed upon scientific methodology.

This is not to say that such unethical data is necessarily irrelevant for science. An argument remains as to whether the findings of such unethical experiments should be used to determine the directions in which it may be reasonable to pursue ethical experimentation. Should the (apparent) success of Milgram’s studies point to the value of studies in authority figure identity and obedience? Should the (seeming) insight of Harlow’s work point to the value of studies of human abusive relationships, and the tendency of the abused to return to the abuser in times of perceived danger, if such studies could be done ethically and humanely? Or would such use of the data itself validate the unethical practices used in collecting the data?

These questions I leave open, for others to ponder further. For now, I am content that we have reached an answer to the central question posed: for we cannot make direct use of data from unethical experiments in science, due to the lack of ethical reproducibility.

American Medical Association (AMA) (1995). ‘Release of Data from Unethical Experiments’, American Medical Association. <>, accessed 29.10.2019

Drummond, Chris. (2017). ‘Reproducible research: a minority opinion’, Journal of Experimental & Theoretical Artificial Intelligence, 30, 1: 1-11

Gray, Peter. (2013). ‘Why Zimbardo’s Prison Experiment Isn’t in My Textbook’, Psychology Today. <>, accessed 23.10.2019.

Haney, C., C. Banks, and P. Zimbardo. (1973). ‘Interpersonal Dynamics in a Simulated Prison’, International Journal of Criminology and Penology, 1: 69–97.

Harlow, Harry F., Robert O. Dodsworth, and Margaret K. Harlow. (1965). ‘Total Social Isolation in Monkeys’, Proceedings of the National Academy of Sciences of the United States of America, 54: 90-97.

Harlow, Harry. (1958). ‘The Nature of Love’, American Psychologist. 13 vol. 12: 673.

Milgram, Stanley. (1963). ‘Behavioral Study of Obedience’, Journal of Abnormal and Social Psychology, 67: 371.

Share on

1 Comment on this post

  1. I really enjoyed this paper. More clearly presented than is usual (good, because none of you have to prove how smart you, we already know) and well deserving of the award (educator from Canada).

Comments are closed.