Over a week in early January 2012, the news feeds of more than 600,000 Facebook users changed subtly: Without users' knowledge, researchers manipulated the feeds' emotional content to examine how Facebook friends' emotions affected one another.
The study on "massive-scale emotional contagion through social networks" (PNASopens in new window, June 17, 2014) generated significant debate in both public and scientific spheres. Much of this debate centered on ethical aspects of the study. In an editorialopens in new window, even the journal's editor-in-chief voiced concern that the "collection of the data by Facebook may have involved practices that were not fully consistent with the principles of obtaining fully informed consent and allowing participants to opt out" (Verma, 2014).
There has been extensive and incisive debate about the ethical and scientific issues arising from the study. Broaddus (2014), for example, noted the lack of transparency as an issue, although the study was ethical under current guidelines for research with human participants, and more important, that ethics (both in public opinion and professional ethics) are embedded in a sense of what is appropriate or wrong, which may change over time, culture and context. Meyer (2014) argued that the sort of research done on the Facebook data was scientifically important, and that we do not want to respond in such a way as to drive such research underground. Wood (2014) reported that a popular online dating site, OKCupid, has engaged in similar manipulation of user-generated site content, with the president of that site observing that "if you use the Internet, you're the subject of hundreds of experiments at any given time, on every site." The issue is not only topical but also involves a relatively common use of Internet data.
One question we can get out of the way quickly: Was the study unethical? It seems not, because any anticipated harms did not outweigh the benefits, and the relevant institutional review board (IRB) determined that it did not fall under its human research program, so the IRB process was properly addressed. What's more, the wording of the terms of agreement on Facebook makes it apparent that "research" may be conducted, although the addition of this wording appears to be relatively recent and may not have been presented to all the users in the study when they joined Facebook.
The debate also raises the question: Could researchers have conducted the study by allowing people to opt out? Probably not, because opting out introduces a potential bias that probably would have swamped the tiny effect size. Of course, we know this only in hindsight. It would also have made the selection process less than random, limiting generalizability.
If the study was not unethical and could not have been conducted another way, why the extensive reaction among the public, the media and scientists? For one, it can be shocking for many to learn that such research has become routine on the Internet. Some of the emotive language in commentaries, which used terms such as "manipulation," "puppet masters who play with the data trails we leave online" and "guinea pigs," underscores that point. Also, the use of the research term "emotional contagion" in the article title was somewhat provocative, since emotions are widely considered to be among the most personal and authentic aspects of one's identity. The fact that emotions may be manipulated might be judged as a more personal attack than manipulation of other aspects of behavior.
Whether such reactions were warranted or not, we need to take seriously the concerns raised. This sort of "manipulation" routinely takes place in advertising research, for example, Broaddus noted. The sense of loss of privacy also remains a core issue. There is a big difference between using a data set to describe a population and relationships between variables and manipulating the behaviors and experiences of individuals within the population. I believe that there is a greater need for care and transparency when an intervention is introduced that manipulates behaviors and experiences.
From a psychological point of view, there are concerns. The fact that such manipulative research is routine in commerce does not obscure the fact that as professionals, psychologists and other behavioral scientists are held to higher standards than, say, used-car salespeople or advertising executives. Nor is the fact that there was little objective harm anticipated a complete justification. There is a distinction between "harm" and "wrong," and what people are reacting to is a sense of "wrong" — both in the lack of transparency and the lack of clear consent.
There are also legal concerns. For one, it's unclear whether the Internet sample contained people under 18 or people from outside the United States who may be subject to different levels of scrutiny in a research study. The European Union, for example, has stringent privacy laws that may have been violated by the Facebook study. It is unclear whether laws governing data protection are the laws in the jurisdiction where research participants reside, where the researchers reside, where the server is located, where the data are analyzed, where the IRB is located or some combination of these. Researchers need to establish this before embarking on Internet studies.
As psychologists and ethicists, what might we advise differently after the fact? One suggestion would be a formal method to opt out of any website research (which would need to be specified as not relating directly to evaluating website services and marketing). That distinction could be made by noting the difference between evaluation for business purposes and research to generate generalizable scientific knowledge — and the fact that the Facebook data were published in a prominent peer-reviewed journal makes the point that this was considered generalizable knowledge in the area of emotion theory. To make this option clear, websites could develop a page in the basic user profile with the title "Consent for general research use" that allows new users to check or un-check a box opting out of research.
It might be preferable to offer a general opt-out on enrollment rather than for a specific research study, since opting out of a specific study introduces a selection bias related to the subject of that study. On the other hand, if there might be a potential harm associated with a particular study, then specific plain-language consent for that study is preferable.
It seems to me that the harm emerging from this study is a sense of a loss of privacy and lack of transparency, or loss of control over what is done with data, and we need to take those reactions seriously. Perhaps it's like undressing in front of an open window and then being outraged that someone watched, but that sense of outrage is real. Perhaps it is also time to revisit our ethical frameworks, which largely date to the predigital age, in the light of recent developments in data collection and storage on the Internet.
Michael W. Ross, PhD, MD, MPH, is a professor of behavioral sciences at the Center for Health Promotion and Prevention Research in the School of Public Health at the University of Texas. He is also a member of the APA Committee on Human Research.
