When we think of research with human subjects, we most often think of medical and sociological research and this is indeed where most of our ethical frameworks stem from.
But medical and sociological researchers are no longer the main groups conducting research with human subjects. 21st-century technology has brought about an absolute explosion in research done with human subjects by those in computer science, technology, and market research. This guide is tailored specifically to developers/market researchers in the private sector conducting tests on users to iterate on their products.
In the 20th century, market research and product development were more of an art than a science and potential harms and ethical considerations were thus much lower. In the 21st century, data on human behavior is abundant and even the most junior developers and market researchers have access to tools and analytics to study human behavior that are far more powerful than the tools even professionally-trained sociologists had access to in the 20th century.
If you’ve been trained in the medical or sociological sciences, you’ve already been exposed to these concepts. If you’ve been trained in any other field, the following 4-minute video from Enhancing Postgraduate Environments is a good primer.
Do not treat people as a means to an end.
Respect their autonomy.
Provide informed consent.
Allow participants to withdraw from your study.
Protect participant data.
Protect people with diminished autonomy.
For example, children, prisoners, and disabled populations. This means you may have to alter the way you deliver informed consent (e.g. providing consent materials to a visually impaired user in the form of audio) to make sure your users truly understand what they are consenting to.
Includes Non-Malfeasance or “do no harm.”
Maximize the possible benefits of research while minimizing possible harms.
For example, you must be willing to terminate a study if, during the course of it, you find the harms to be too great to either the test users or society.
Select research subjects fairly.
Compensate them for harm done.
Share the benefits of research.
For example, translate your research findings into the native language of the population you studied so they can benefit as well.
If research with human subjects includes studying any aspect of human functioning or behavior, it is clear that many developers, market researchers, and computer science researchers are engaged in research with human subjects on a daily basis whether they realize it or not.
But how can we apply these principles to such settings?
Imagine you are starting a software company. You plan to use mixpanel for metrics and do a lot of A/B testing to understand the human behavior of your users but you want to do it in an ethical way. Let’s run through each key principle and apply it to this situation.
Respect for persons includes respecting autonomy and privacy and protecting vulnerable populations. In order to meet these requirements, you could:
Provide easy-to-understand terms of service that truly provide informed consent.
Inform users of all the data you are collecting from them and how it will be used.
Inform users about the possibility of A/B testing and allow them to opt-out of being a part of those experiments.
Make sure you are following best practices on the protection of user data.
Take into consideration vulnerable populations that may need additional layers of informed consent.
Beneficence includes not harming your users and maximizing the benefits of your research. If your product provides benefit to users and the purpose of your A/B testing and mixpanel metrics analysis is to increase the benefit to your users, that’s beneficence.
A company that does not follow the principles of beneficence or respect for persons would be a company that knows its products are harmful to users but nevertheless experiments on them in order to find ways to manipulate their behavior to increase harmful usage, using users as a simple means to an end (profit).
A recent example of a company that purported beneficence as it’s reason for being then preyed on a vulnerable population (children) without much regard to respect for persons or implementing non-malfeasance is Juul.
So long as your products provide real benefit to users, the purpose of your research is to increase the benefit to your users, and you are willing to terminate experiments that appear harmful to either your users or society, you are following the principle of beneficence in practice.
Justice includes selecting participants fairly, compensating for harm done, and giving people what they deserve/sharing benefits.
This means when you are selecting which users to test on, you should try to get a representative sample and not just select the easiest populations which are often the most vulnerable populations. For example, if you can only do a test on users living in a specific country with extremely lax privacy laws, it is unfair to continually exploit the vulnerability of that population for the purpose of changing your product for other populations.
Though you might not mean harm, especially if you are following these principles, sometimes harm is accidentally done during the course of an experiment (a reason you must be willing to terminate an experiment if unforeseen harm arises).
In the least extreme scenario, an experiment might cause a minor inconvenience to a user and you could make up for that with a heartfelt apology, discount, or refund. In this context, justice looks like good customer service.
In the most extreme scenario, an experiment might cause major harm to a person’s health or wellbeing and if it’s systemic, could result in a well-deserved class-action lawsuit. If you are following the principles set out in this guide, you will catch harm done to users, stop it, and make amends for it before it can escalate to such a point.