By Tim Rubbelke
In
mid-2014, researchers at Cornell University released a study titled
“Experimental evidence of massive scale emotional contagion through social
networks.” The general idea of the study
was that manipulating the Facebook feeds of users to display either primarily
positive or negative posts caused subjects’ mood to shift accordingly.[1] The study involved over 600,000 individuals.
Soon
after the study was published, tech websites took notice.[2]
One of the obvious questions was whether
the study participants had consented and whether they knew they were part of
this study.
Facebook
pointed to a broad clause in its “Data Use Policy” (DUP) as implying consent
(although it seems no one in the study was aware of their participation.) Some sites questioned whether the clause in
question was even in the DUP at the start of the experiment.[3]
In
light of the criticism, the Proceedings of the National Academy of Sciences (PNAS) (the journal that published the research) issued
an Expression of Concern about the study.[4] In it they noted that Cornell’s IRB chose not
to weigh in on the research because the data their researchers received was
anonymized and thus it was not human subjects research under the regulations.
While
Cornell’s decision likely fits the technical description of the human subjects
research regulations, it serves to highlight a very interesting question: when
researchers (and institutions) receive data, should they be concerned about how
the data was collected?
One
hopes that Facebook carried out the experiment in an ethical manner, but
without oversight we cannot be sure. It
certainly seems that at the very least many participants were subjected to a
form of harm—by having their moods unknowingly altered to be sadder. Perhaps more disturbingly, the study gives no
indication of what would happen if a participant’s emotions swung far beyond
the expected range.[5]
The
rules and regulations surrounding academic research are designed to hold it to
a high ethical standard, especially regarding the treatment of human
subjects. When data is shared between
academic institutions then the research will be subject to the oversight of at
least one institution. In the case of
data from private sources (such as Facebook) things are substantially murkier. As the PNAS expression of concern noted,
privately funded organizations like Facebook do not have to follow the Common
Rule. This is compounded by the fact
that, as saying on the internet goes, “If you’re not paying for it, you are the
product.”[6]
Still, as one expert interviewed by CNet pointed out, the experiment carried
out by Facebook went beyond the normal types of platform and advertising
enhancement most websites carry out.[7]
Furthermore if academic researchers are going to accept data from privately
funded entities, and wish to maintain the high ethical standards, they should
be cognizant of how the data was collected.
As
Kahn, et al. note, it’s inadvisable to try and shoehorn a 20th
century research regulation apparatus onto a 21st century world.[8] And yet, as we try to bring the research
oversight into the modern world we should be careful not to discard bedrock
principles either. After all, public trust in academia is what fosters
participation in research.
Tim Rubbelke is a PhD Candidate at the Saint Louis University Albert Gnaegi Center for Health Care Ethics. He contributes regular pieces on research ethics.
[1] http://www.pnas.org/content/111/24/8788.full.pdf
[3] http://www.cnet.com/news/facebooks-emotion-manipulation-study-faces-added-scrutiny/
[5]
Additionally there’s no information about the age of the subjects.
[6] http://lifehacker.com/5697167/if-youre-not-paying-for-it-youre-the-product
[8] http://www.pnas.org/content/111/38/13677.full
