In his book Who Goes First, Lawrence K. Altman narrates the rich history of self-experiments in medicine. Self-experiments in medicine have many advantages over experiments done on other people. Arguably, the greatest one is that they align personal incentives with social incentives. As a society, we want to come up with life-saving medicines and other cures while not exposing research participants to undue risks. Researchers and doctors have both intrinsic and extrinsic motivation to strike the right balance when working with volunteers. Nonetheless, experimenting on oneself creates the strongest form of motivation to get things right. After all, if something goes wrong in a self-experiment, the self-experimenter might die or suffer permanent disability.
If self-experiments create good incentives in medicine, they should hold similar potential in other fields as well. Consider the social sciences, and more specifically, psychology. While there are some examples of rigorous self-experiments in psychology (e.g., Hermann Ebbinghaus’s research on memory), they aren’t as important or prevalent as those in medicine.
In this post, I summarize my current thinking on some aspects of self-experiments in the social sciences.
Social science self-experiments in everyday life
Regular people already do occasional self-experiments. Sometimes they do it consciously, especially when the goal of the self-experimentation is increasing productivity or efficiency. In fact, the entire industry of life hacks and self-help books is built on the idea of self-experimentation. Outside these domains, however, self-experiments are much less prevalent or deliberate. If people do a self-experiment, it’s typically because they stumble upon it by chance.
Imagine a couple who always talk about shared interests late at night, simply because that is when they are free. One day, their morning plans are canceled, and the conversation about shared interests happens in the morning. It’s even more enjoyable than usual. They decide to have these types of conversations only in the morning from then on. Albeit imperfectly designed, they are doing a self-experiment on the effect of the time of day on conversation quality.
The above examples are from people’s personal lives, but their professional lives are no different. If anything, most people prioritize improving at their jobs over, say, building better relationships with family members.
Journalists and politicians as closet self-experimenters
A group of professionals involved in self-experiments are journalists. Consider the journalist who does immersive reporting about water scarcity in the Sahel. He lives for two weeks with a local family, which includes walking 10 km every day to fetch water from a well. He may frame his mission differently, but part of his job involves a self-experiment on the effects of water scarcity on people’s day-to-day happiness (and other life outcomes). It’s one thing to read in a study that people’s self-reported happiness is 0.72 points lower in that region compared to another region with better access to water. It’s another to have to traverse a harsh terrain day after day regardless of how you feel or what else you would like to do with your time.
Famed biographer Robert Caro temporarily moved to rural Texas to immerse himself in the culture that shaped his research subject, US President Lyndon B. Johnson. An extreme form of self-experiment for sure, but one that must have contributed to his ability to create a vivid portrayal of his subject.
Other professionals may have less noble goals when doing self-experiments. Imagine the middle-aged politician who tries to gain popularity by portraying himself as a devout family man while using a secret bank account to pay for his teenage lover’s lavish lifestyle. Even though his opponents may frame his actions differently, he is conducting a self-experiment. He investigates the effects of presenting oneself in a socially desirable (vs. undesirable) way on political popularity.
Social science self-experiments in research
In most of the examples I mentioned so far, self-experimentation is not very rigorous. The point is rarely to find out something scientifically interesting. A researcher might want to understand how presenting oneself favorably (vs. unfavorably) influences a politician’s popularity. But the politician? He is merely interested in gaining or retaining power. It seems rather futile to ask him to be in the control condition—being upfront about his adulterous and dishonest life.
Rigorous self-experiments seem rare in the social sciences. This belief is difficult to test, as self-experiments may go unmentioned. A researcher might get an idea from a self-experiment and then tests that idea on a sample of participants. When writing up his findings, he considers the source of his idea irrelevant. This way, the self-experiment might never enter the scientific record. Should this pattern repeat itself consistently, we might find ourselves in a situation where self-experiments are widespread even though we believe them to be rare.
Let’s assume that social science self-experiments don’t suffer from this reporting issue. In that case, what are some possible explanations that could be behind the relative scarcity of social science self-experiments?
The first candidate is that social science self-experiments are not that useful because it’s fairly easy to find third-party study participants. Contrast this with the case in medicine, where getting volunteers might be very difficult. For example, Dr. Frederick Prescott wanted to explore the potential of a substance for general anesthesia. The substance happened to be curare, which is the paralyzing agent used by indigenous peoples in South and Central America to kill their prey. Well, good luck signing up your friends for the curare experiment. In contrast, clicking through some pages or doing some trivial tasks on a university campus is a much more palatable proposal to prospective participants.
So, social science researchers may do less self-experimenting because they have easy access to third-party participants. This consideration aside, the marginal value of an extra participant is lower in the social sciences than in medicine.
The typical effect size in the social sciences is smaller than in medicine. Showing a large effect requires fewer participants. In extreme cases in medicine, even a single participant can provide credible evidence for a hypothesis (e.g., to establish channels of transmission). In the social sciences, a single participant can at best give us some ideas that might be worth investigating.
A different angle on the same topic is variability. In a famous self-experiment, German doctor Werner Forssmann inserted a catheter into a vein of his arm in an effort to reach his heart. (The self-experiment was successful, and he survived.) Whether it’s a young German doctor in 1929, a Mongolian housewife today, or a Chilean bricklayer in 50 years, the catheter will follow the same itinerary to the heart.
Now imagine you stop every passerby in your neighborhood and ask them politely if you can make a small incision at their elbow crease and push a catheter through their vein into their heart. For extra effect, perhaps even wave a 65-cm-long catheter before their eyes. Some of them may call the police, others may call an ambulance, and others may look for the hidden camera. The point is that the social reactions to this request will be much more variable than the biological reactions to catheter insertion.
Beyond differences in the marginal value of an extra participant, there is also a difference in the type of question that can be studied with a self-experiment. In medicine, with few exceptions, what can be studied on someone else can also be studied on oneself. Not so in the social sciences. For example, you can study the effects of forced relocation on a person's life satisfaction. History is, unfortunately, full of these types of experiments, more commonly called population resettlements or deportations. But forcibly relocating yourself is an oxymoron, akin to being voluntarily drafted into an army. Similarly, studying your cognitive blind spot via self-experiments is difficult, as you cannot study something you are not aware of.
The case for social science self-experiments
While the picture I painted of social science self-experiments might seem somewhat grim, I think they are, in fact, quite useful. This is because doing a self-experiment creates indirect benefits on top of the direct benefits, and their cost is typically low.
One indirect benefit of social science self-experiments is that they help one develop an appreciation for variability. Someone could reasonably object that people already know that their fellow citizens react to things differently. Sure, people know this in the abstract. The crucial question, however, is whether they can apply this knowledge to a new situation. They realize that preferences within some groups (e.g., their families) vary, yet they may fail to apply this insight when thinking about the preferences of outgroups, such as the citizens of Vanuatu or care home residents. Knowledge transfer across contexts is difficult, but seeing variability in one context makes it marginally more likely that people will recognize variability in another context. An easy way to experience more variability is by doing lots of self-experiments.
A second benefit of self-experiments is that one can experience things he wouldn’t experience as a neutral observer. This is useful because we frequently misjudge how we would feel in a given situation: We struggle to imagine how bored we would feel punching numbers into a database or how sad we would feel when our friend moves to a new town. Developing more empathy for study participants, fixing issues in the experimental design, and generating new research ideas are some of the other indirect benefits of doing self-experiments.
Considering that many social science self-experiments are cheap and easy to do, many will pass a simple cost-benefit analysis. It would be nice if others did more social science self-experiments. But if a post about self-experiments should teach us something, it’s that we should go first more often. In this spirit, I plan to do more self-experiments over the coming months and share some of my findings here.