Ethical considerations
Last updated
Last updated
Part of the data analyzed by computational biologists is derived from studies performed on human subjects. Thus, even computational scientists need to be aware of the ethical considerations that surround human subjects research. More broadly, however, computational research itself, even if it doesn't involve medical applications, may impact individuals. The ability to carefully consider ethical concerns is, therefore, important for all computational scientists.
A better understanding of ethical thinking is particularly important for those who have a "hard sciences" background. As computer scientists, engineers, mathematicians, etc., we are used to the rigor provided by objective data, mathematical theorems, and widely-accepted scientific theories. Ethics appears to us as too "fuzzy" and indetermined. One's opinion or personal feelings on a subject seem to be more important than any universally-accepted "truths". The point of this section is to demonstrate that frameworks exist that can bring rigor to the assessment of ethical considerations.
Before proceeding, it is important to stress that what follows is not an attempt to tell the reader what values they should have. First and foremost, this section is meant to introduce you to ethical thinking, largely viewed through the lens of human subjects research. As defined at :
Ethical thinking is the process of identifying and describing ethical issues in a variety of contexts, articulating the ethical considerations involved in different responses to those issues, and providing a rationale for a position that addresses those considerations.
Second, this section will provide you with a high-level understanding of the legal/regulatory factors that govern ethical considerations in human subjects research. To some extent, laws and regulations define the high-level values that need to be observed when making ethical decisions. The law helps provide rigor in evaluating the right/wrong dichotomy (within the narrow context in which specific laws/regulations apply).
In the United States, federal policies surrounding human subjects research were strongly influenced by the , a 1979 document created by the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research that outlines the ethical principles that must underlie human subjects research. The Belmont report was a response to several major and public failures of ethical decision-making, such as the medical atrocities committed by the , or the study.
Before you proceed, please read the .
While, for many of us, the words "human subjects research" conjure images of doctors poking patients with needles, this concept covers a much broader range of research activities, including testing the usability of software or websites, trying out new methods of teaching, and many more. Recent examples of alleged ethical failures in non-medical human subjects research include studies conducted by the dating site OKCupid (this provides an overview), and emotional contagion studies conducted by Facebook (this provides an overview). In 2012, the principles of the Belmont Report were revisited for the specific context of Information and Communications Technologies, leading to the creation of the (also see the corresponding ).
In the United States, human subjects research is regulated through the - a broad federal policy that is implemented in agency-specific regulations by the agencies that fund human subjects research. Importantly, from a regulatory standpoint, these regulations only apply to entities that accept federal funding from the agencies that are signatories to the Common Rule. Thus, what a company does on their own dime is not regulated through this mechanism.
Before proceeding, it is important to make the distinction between research and practice. This distinction is important from a regulatory/legal point of view. The Common Rule regulations apply only to research activities. For example, a doctor trying out different antibiotics to cure a patient's strep throat would not be subject to the Common Rule, but a doctor performing a clinical trial to determine which antibiotic is most effective against strep throat would be. That doesn't mean that a practitioner is not subject to ethical/regulatory considerations, just that it is a different set of regulations than those spelled out by the Common Rule (e.g., medical practice is governed by the ).
The Belmont report makes the distinction between practice and research as follows:
Practice: Interventions that are designed solely to enhance the well-being of an individual patient or client and that have a reasonable expectation of success.
Research: Activity designed to test an hypothesis, permit conclusions to be drawn, and thereby to develop or contribute to generalizable knowledge (expressed, for example, in theories, principles, and statements of relationships).
Research is usually described in a formal protocol that sets forth an objective and a set of procedures designed to reach that objective.
When a clinician departs in a significant way from standard or accepted practice, the innovation does not, in and of itself, constitute research. The fact that a procedure is "experimental," in the sense of new, untested or different, does not automatically place it in the category of research.
The Belmont report sets out several broad ethical principles that need to be taken into account when making ethical decisions regarding human subjects research. These are described in more detail below.
This principle states that an individual's autonomy needs to be respected, and that individuals with impaired autonomy must be protected. As a concrete example, before enrolling someone in a research study, they have to be provided with sufficient information (and time) so that they can decide whether they want to be involved in the study, and their wish must be observed. This is related to the concept of informed consent — one must be informed and consent to their participation.
As an example of individuals with impaired autonomy, studies on children require careful protections to be put in place because children may have neither the power, nor the knowledge to be able to become informed and to consent to their participation in a study. Other examples include situations in which the researchers have power over the individuals enrolled in their study (e.g., a professor enrolling students from a graduate program the professor directs).
It is also possible that certain studies cannot be performed if the participants know they are enrolled in the study (e.g., a deception study where some participants are lied to intentionally). Again, since participants cannot consent, upfront, to participate in the study, the researchers must carefully protect them from harm. In cases such as this, when informed consent cannot be received prior to participation in the study, researchers must inform the participants and obtain consent as soon as practically possible.
This principle states that researchers must take into account the subjects' well-being. This principle is aligned with the Hippocratic Oath's "do not harm" principle. Beneficience also involves weighing against each other the benefits and risks that could arise from a subject's participation in a study.
The benefits are often considered in terms broader than an individual — would the successful conclusion of a research project provide benefits to society or to a broader group of individuals. Say, for example, that a study is testing a new drug meant to combat morning sickness — a common condition in early pregnancy — and that a male volunteer agrees to take the drug to confirm it is safe. There is no direct benefit to the individual participant since they are not expected to experience morning sickness, thus, for the individual participant this study only has risks (e.g., the risk that the drug has unintended negative side-effects). This risk, however, is weighed against the benefit of providing pregnant women with relief from this condition.
The principle of justice balances two questions: (i) Who benefits from the study? (ii) Who bears the burdens of the study? Coming back to the example of the morning sickness drug above — if the study only enrolled men, then the principle of justice would be violated. The study, if successful, would benefit women, but the risks are all assumed by men.
The groups implicit in the word "who" vary depending on context, and must be carefully considered in the context of applying the principle of justice. As an example, the Belmont report specifically calls out the Tuskeegee syphilis study, where the participants were all rural Black men, even though the disease was not limited (or even most prevalent) in this sub-population.
When considering justice, it is important for researchers to ensure that study participants are selected because they are representative of the population that will benefit from the results of the study, and not simply due to convenience (e.g., because they are poor and, thus, more likely to be willing to accept the study's financial compensation, or because they live near the study site, etc.).
The Menlo Report adds a fourth principle to highlight the need to conduct information and communication technology research in a way that is compliant with existing laws, transparent, and accountable. The Compliance component of this principle asks that research be conducted in a way that obeys relevant laws, under the assumption that the legal framework ensures that the research respects public interest. In case different laws conflict with each other or are considered to contravene public interest, this principles argues that researchers must provide an ethically defensible justification for deviating from the law and that they be willing to accept the consequences. The Transparency and Accountability component of this principle argues that the researchers transparently communicate the goals, the methods of the research, and the associate risk in order to engender trust and to hold the researchers accountable for their actions.
If you take a bit of time to reflect, you will realize that the different principles conflict with each other. For example, when testing medication for a common childhood illness, we may conduct studies on adults since children have limited autonomy, and we would, thus, violate the principle of respect for persons. But doing so, we would violate the principle of justice, since the study participants do not belong to the group that benefits from the study. In general, you will find that ethical questions rarely have clean answers. How do we, then, make decisions?
First and foremost, it is useful to carefully view the proposed research through the lens of each of the three principles, and to highlight areas of concern. Doing so forces us to think through the implications of each of the ethical principles before proceeding, but we are still left with the difficult task of balancing out the conflicting requirements of the ethical principles.
It is important to recognize that no individual can (nor should) make this determination. A diversity of perspectives is valuable when discussing complex ethical considerations in order to make sure that the ethical principles are adequately balanced against each other and that the study best serves the interests of individuals and society, more generally. For this reason, the Common Rule mandates the creation of Institutional Review Boards (IRBs) — committees that evaluate proposed research according to the framework provided by the Belmont Report. An IRB must have at least five members and its membership must be diverse across gender, racial, ethnic and other dimensions. The Common Rule also mandates that at least one member of an IRB must be a scientific expert, and at least one member must have expertise in non-scientific areas. Finally, one of the members of an IRB must not be affiliate with the institution convening the IRB in order to bring an external perspective. The Common Rule also allows/encourages the addition to an IRB of individuals that can bring specific expertise that may be needed to evaluate a particular proposal.
In short, the Institutional Review Boards mandated by the Common Rule are designed to leverage a diversity of expertise, opinions, and lived experiences when making ethical decisions about human subjects research, in hopes that the consensus opinion will best protect the interests of study participants and of society at large.
The discussion so far has focused on the relatively narrow scope of human subjects research that is subject to the requirements of the US federal Common Rule. The reader may reside in a different country, or engage in activities that are not subject to the Common Rule regulations, thus we wanted to provide some recommendations for ethical decision making in a broader context.
Professional societies or disciplinary governing bodies are likely to have developed ethical decision-making frameworks similar to the Common Rule that are specific to the unique circumstances of the individual discipline or society. Please educate yourself about the ethical resources available in your country, discipline, or domain of work. The process for applying such principles that we have outlined earlier can be used within other decision frameworks. Most importantly, it is critical that difficult ethical decisions be made only after carefully consulting a diverse set of opinions. Formal or informal review boards should be part of your regular practice.
Decision-making frameworks are also developed by social scientists. While we, as computational scientists, may deride the "softness" of social science, we have to acknowledge that it is, ultimately, a science, no less valuable than others. The conceptual frameworks developed by social scientists provide us with a structured way of thinking about or discussing complex concepts in both research and practice, and at a broader societal level. We encourage the readers to educate themselves about prevalent social science conceptual frameworks. Practicing the use of such frameworks to guide our thinking about a wide range of matters will help us be better citizens of this world.
Scientists have developed a vaccine for a rare but severe illness. Animal studies and human safety studies (in uninfected individuals) have demonstrated that the vaccine prevents the illness in animals, and that it is safe to administer to humans. Now the scientists want to verify whether the vaccine is effective in humans, i.e., whether it can prevent infection with this severe illness. Since the illness is fairly rare, they cannot vaccinate a group of people and hope that some of them will be exposed to the virus causing the illness (which is a common strategy for prevalent illnesses, such as the flu) — they would simply have to enroll too many people in the study to be able to observe enough exposures to the virus to be able to draw robust conclusions about the effectiveness of the vaccine. Instead, they propose to seek volunteers who agree to be intentionally infected with the virus (this is called a challenge study). Because the illness is severe, all study participants will be admitted to the hospital for 2 weeks (accounting for the typical duration of the illness and recovery period). During this time, study participants will receive free meals and medical care as well as a stipend of $200 per day. The risks and benefits of this study will be carefully communicated to study participants, and they will be allowed to voluntarily choose whether they want to participate, as well as whether they want to quit the study at any point in time.
Should this study be allowed to proceed?
Let us look at each of the ethical principles:
Respect for persons: The description above indicates that participants will be informed about risks/benefits and they will be able to voluntarily choose whether to participate. Thus, the principle of respect for persons seems to be satisfied. However, some information is missing - the description does not indicate whether information for study participants will be available in multiple languages. Medical terminology may be difficult to understand by individuals who do not have English as a first language, thus they may not be sufficiently informed to consent to participating. An IRB would typically request to see the plan for how participants will be consented for participating in the study, including the specific hand-outs that will be provided, to make sure that they are written clearly enough for a lay person to be able to understand. Furthermore, it is important to consider who may enroll in such a study. For many working professionals, $200 per day is not sufficient compensation for spending 2 weeks isolated in a hospital bed. Thus, study participants may predominantly be individuals with a precarious financial situation. Thus, they may not join the study freely (as mandated by the Respect for Persons principle) but out of necessity.
Beneficence: The description indicates that the illness targeted by the vaccine is a severe disease, thus there appears to be a societal benefit that will result from this study. But what is the benefit/risk ratio for the study participants? The researchers have taken steps to mitigate the risk — they made sure the patients will be in an environment where they can be treated if they fall ill. But the description does not provide sufficient information about the actual risks. Is this an illness that can be effectively managed in a hospital setting, or one that has a high mortality rate despite medical care? An IRB would need to receive further information before determining that the risk is worth given the potential benefit. Importantly, even if there is a high risk of death of permanent harm to individual patients, studies may still proceed if the societal benefits are significant. In the early days of the COVID pandemic, when treatments were not yet available for this disease, many individuals volunteered for SARS-CoV-2 challenge studies, knowingly accepting the risk of death in order to help accelerate the development of vaccines that will eventually save millions of lives.
Justice: The description of the study does not say anything about the groups most affected by the illness or the recruitment strategy for the study. This information will have to be provided before deciding whether the principle of Justice is observed. The comment made earlier about the socioeconomic status of individuals who may participate in the study does suggest concerns. While the study may benefit society as a whole, it's possible that this study would only burden individuals of a lower socioeconomic status, thus it would not be considered just.
Conclusion: So...should the study proceed? That's for you to determine. Grab a group of friends and discuss. Do they agree with your assessment? Do they manage to change your mind?
Facebook conducted a study to evaluate whether the content that users see impacts their mood. In this study, the feed recommendation algorithm was modified to provide some users with content determined by an algorithm to be positive, and other users with negative content. Researchers then looked at how the "mood" of the content posted by users changed in response to the content they saw in their feed. The users were not informed of this study, either before participating, or after the conclusion of the study. Facebook partnered with researchers from a major university to analyze the resulting data and the results of the study were published in a major journal.
Was this study conducted ethically according to the principles described in the Common Rule?
This example is a bit different than the previous one, in that we only found out about it after it had concluded. Thus, the question here is not whether the study should be allowed to proceed, but whether the study violated ethical principles. Let us walk through the different principles.
Respect for persons: As indicated in the description of the study, participants were not informed that they were part of the study and were not allowed to choose whether or not they wanted to participate. Their autonomy was clearly not respected, therefore the study does not meet the expectations of this ethical principle.
Beneficence: For individual participants the description provided does not allow us to determine a potential benefit, or the risks associated with participating in the study. One could see a potential benefit to participants who received positive content. One could also make the case that a more thoughtful feed recommendation algorithm could benefit Facebook users more broadly. This broader benefit is unclear since no information was provided about the ultimate goal of the study — was this simply curiosity, or will it have resulted in changes to the recommendation algorithm that ensured users received balanced content? Regarding risks, psychologists have argued that providing some participants with a stream of negative content could trigger or exacerbate mental health conditions, and that, therefore, the risk entailed by this study was unacceptable.
Justice: There is simply no information in the description above that indicates whether any particular group of Facebook users was targeted. Assuming that participant selection was random, we can possibly infer that the population benefiting from the study (Facebook users in general) and the population involved in the study were similar, thus the principle of Justice appears to be satisfied.
Do ethical considerations even apply?: This study is pretty clearly research because the conclusions of the study were published in a scientific journal (the implication being that the study led to generalizable knowledge). It also clearly involves human subjects, and falls in the category if behavioral research, which is explicitly targeted by the Common Rule. Facebook, however, argued that the Common Rule doesn't apply since they are a corporation that has not received any federal funding for this research. Furthermore, to counter the claim that the study violates the Respect for Persons principle, they argue that the study participants implicitly consented to the study (and any other study the company may want to conduct) by agreeing to the Facebook terms and conditions. The university researchers, were, however, recipients of federal funding, and therefore the Common Rule would apply to their work. They have argued, however, that they simply analyzed data that had already been collected, thus their work did not meet the criteria for behavioral human subjects research.
Conclusion: Was this study conducted ethically? Does the Common Rule apply? What do you think? How about a group of your peers after you carefully discuss the study?
Researchers are planning to develop new facial recognition software. As training data for their software they want to download data from public data set of celebrity photos. A recent news story indicated that several celebrities had objected to their image being used to train machine learning systems, though the data continued to be publicly available under licensing terms that did not prohibit the use of the data.
Should the researchers be allowed to conduct this study?
Respect for Persons: The description above indicates that some of the individuals whose data will be used have said they do not want their data used in this way. If the researchers proceed without honoring this wish, then the principle of Respect for Persons is violated. Note that, legally, celebrities do not have the same right to control their image as "normal" people do. But here we are not asking a legal question, rather an ethical one. Law and ethics do not necessarily align, and the extent to which they do varies from country to country.
Beneficence: The description says nothing about the ways in which the facial recognition software will be used, thus it is not possible to assess the potential benefit to the celebrities involved or society more broadly. Perhaps better facial recognition for celebrities would make it easier for them to access events (e.g., the Oscars) without having to carry an ID or ticket. Such a benefit may offset the potential risks involved by training the facial recognition software on their images. How about the risk? It appears fairly low — how can one be inconvenienced by someone using their public image in training a machine learning algorithm. But what if this software is designed to be used by border control officials? Perhaps including a celebrity's image in the training data may cause the system to inadvertently flag that celebrity when they try to travel. Or, perhaps, a celebrity may be incorrectly implicated in the conduct of a crime as police use the system to review security camera footage. Researchers should consider such potential risks carefully and explain why such concerns are not relevant here, or how such concerns get balanced against the overall benefits of the study. An IRB reviewing this study should ask for such clarifications.
Justice: Assessing the principle of Justice is also hard to do given the information provided. If the facial recognition system will only be used for celebrities, then Justice appears to be respected since the participants and the eventual population benefiting from the study are the same. What if the system will be used in the general public? Celebrities are definitely not "typical" individuals — they tend to be more attractive, richer, and also the group "celebrity" is less diverse than the general population. In this case, Justice would not be respected since the group assuming the risk of the study is different from the group intended to benefit from it.
Do ethical considerations even apply? Is this even human subjects research? The question of whether this is research is unclear given what is presented. If the researchers intend to publish papers that describe knowledge applicable beyond the scope of this study, then yes, this is research. If, however, they are simply using the celebrity images to improve the performance of a specific facial recognition system, this may be viewed as a form of practice, thus not subject to the Common Rule. Are human subjects even involved? The researchers did not enroll the celebrities in the study and did not take their pictures. All they did was reuse data that was publicly available. IRBs would probably say that human subjects protections do not apply here. Should the regulations change to include situations such as the one in this case study? This is a question that the computer science community is increasingly considering as machine learning algorithms are being trained on vast volumes of public data.
Conclusion: What do you think? What do your peers think? Do your or your friends' opinions change after debating this study?
The further states that:
A few examples may be useful to highlight how particular cases may be evaluated from an ethical standpoint. In addition, several resources exist in the context of information and communication technology research to help and explore considerations related to data use - .