Saturday, May 14, 2016

More on the ethics of the OKC/OSF data dump

Here's another take on what was ethically incorrect about the Emil Kirkegaard data dump of nearly 70,000 OKCupid users.

There are essentially four points that are to be considered:

1. OKC users did not consent to their data being used in the way Kirkegaard used them or to be shared. In fact, even a cursory reading of the terms of service at OKC are clear on the matter of how their data are to be used.

2. OKC itself did not consent to the data being used as it was. We can certainly quibble about concern over a company's well-being, but certainly there is precedent for researchers obtaining a company's permission to analyze their members' data (with safeguards for anonymity put in place). The reason I add that last sentence is that some of my first research experiences as an undergrad involved analyzing data from a computer matching service, from which my mentor had obtained all necessary permission. Everything was above board. In the process, we were able to make a number of statements with that data, including confirming that successfully matched couples tended to have considerably more in common than couples that were not successful matches. It made for a couple conference presentations, if nothing else, and no one's privacy was harmed in the process. A win-win in my book.

3. Which leads to a third point - the way Kirkegaard went about releasing the database could lead to real harm to real people. That is not how we operate as scientists. A guiding ethical principle that was imparted to me and that I share with my students is that we do no harm to those we study. Putting others' privacy at risk is a good reason to not make a database public or go forward with a particular project. The potential for individuals in the OKCupid database to be personally identified is one I find rather unsettling (and that is really an understatement), especially to the extent that their being personally identified could lead to real physical or financial harm. The publishers of the database showed cavalier disregard to that possibility, and those who are using this data going forward are doing likewise.

4. Any work involving human participants requires review by an institution's ethics panel. That applies not only to faculty and staff members of an institution, but its students as well. Students in particular are still beginning to learn the research process, and the idea of students (at any level) being turned loose to simply run whatever they want without oversight is not something a reputable institution will stand for. However, regardless of whether one is a student or a seasoned professional, and regardless of how excited one might be about their research ideas, there is a need for an impartial third party to provide some modicum of oversight - if nothing else to determine the ethical soundness of a particular project. In addition, the Kirkegaard data dump along with its accompanying paper provides another problem: conflict of interest. Not only did Kirkegaard entirely sidestep his institution's IRB, but he made sure that another potential safeguard was sidestepped: the role of the journal editor and peer reviewers. As the editor in chief of the "journal" in which his work was published, Kirkegaard is his own judge, jury, and executioner.

I am sure we all have our own various war stories to tell about our institutions' IRBs or ethics panels, but at the end of the day, they do serve an important function. Regardless of any frustrations I've had with my own, ultimately I and my students get to conduct our research. We may a wait on our hands, but we'd rather err on the side of safety.

I am sure that this incident will cast a shadow over the open science movement, which is a shame since I do think that this is a movement that serves a purpose as well. The appeal of the open science movement was its apparent dedication to transparency, leading to better and more ethically sound science. OSF, which housed the data set in question, should simply remove the data set posthaste. In other words, an organization that is the face of the movement needs to take a strong stand now, rather than freeze in a time of crisis. In any form of communication, there is a fine line between openness and TMI (aka., too much information). Kirkegaard has erred on the side of TMI, and could drag down a potentially beneficial movement within the sciences with him in the process. That should concern all of us.

Michael Gerson on dehumanization

Here is a clip (for those of us who would otherwise be frustrated by columns that are behind WaPo's  paywall:

“A significant part of the Roma are unfit for coexistence. . . . These Roma are animals, and they behave like animals. . . . Inarticulate sounds pour out of their bestial skulls. . . . These animals shouldn’t be allowed to exist. In no way. That needs to be solved — immediately and regardless of the method.”

These words were written not in 1943 but in 2013. They are from an article by Zsolt Bayer, one of the founders of Hungary’s ruling party, Fidesz, and a friend of Prime Minister Viktor Orban. Bayer’s attitude toward the Roma (Gypsies) is part of a broader theory that big corporations, leftists, Jews and Muslim migrants are engaged in a conspiracy to undermine Hungarian identity. “There are all kinds of weapons: traditional, chemical, atomic,” Bayer argues. “And now we see that there are also racial weapons. This is the weapon that they, the ‘invisible hands,’ have employed against Europe and against the white race.”

Orban is not quite so blunt, but he seems more than willing to gather the political benefit of ethno-nationalism. “We, the Hungarians of national solidarity,” he has said, “must squeeze all disunity out of Hungarian life.”

Hostility to outsiders, of course, preexisted the political movement taking advantage of it. But what role does leadership play in encouraging this attitude? This has been a topic of recent research by Emile Bruneau of the University of Pennsylvania and Nour Kteily of Northwestern University. They have devised an appropriately offensive scale on which to measure blatant dehumanization. In September 2014, a representative sample of Hungarians was asked to place Muslim migrants somewhere on the familiar “ascent of man” scientific illustration — the one showing the gradual development from ape to Homo sapiens. The same survey was conducted in October 2015. In a little over a year, the level of blatant dehumanization in Hungary doubled.

There are a number of possible explanations. But Bruneau postulates that political rhetoric played a role. “When people see this as normative,” he told me, “they are more likely to express themselves.”

Bruneau has also studied the disturbing neuroscience of bigotry. One might expect dehumanization to light up emotional, pre-rational parts of the limbic system. Instead, he said, “it is deeply seated in the cortex, in a reasoned cognitive response.” Viewing others as less than human involves a very conscious and deliberate decision.

“Dehumanization,” argued Bruneau, “morally disengages us.” Most humans hold to a morality that forbids harm to other humans. But if someone is regarded as less than human, those moral rules no longer apply. This rationalization is what allows people who commit genocide to go home, kiss their children and sleep at night. It is also what leads Bayer to say: “Whoever runs over a Gypsy child is acting correctly if he gives no thought to stopping and steps hard on the accelerator.”

How does this relate to U.S. politics? In a survey of Americans conducted by Bruneau and Kteily, the dehumanization of Muslims (as you’d expect) was a strong predictor of support for policies such as carpet bombing in the Middle East and denying visas to Muslims. “Conservatism does predict some support for these positions,” said Bruneau, “but dehumanization goes above and beyond this. It is more strongly predictive than political ideology.”

Blatant dehumanization was also more strongly correlated with support for Donald Trump than for any other candidate.

This recent research seems to dovetail with quite a bit of what we know about the impact of dehumanization, including attitudes toward torture, where a student and I were able to demonstrate that attitudes toward torture were significantly more favorable when the victim was Muslim than when non-Muslim American.

Thursday, May 12, 2016

Ethically challenged "researcher" potentially doxes OKCupid website users

Since I had written a bit about the problems with researchers analyzing data from social media platforms without getting prior informed consent (remember Facebook?) I thought this latest fiasco deserves to at least be highlighted. I noticed a mention of the ethical breach in question in my twitter feed (via Vox) as I was taking a break from grading. Oliver Keyes has a well-written take-down of the ethical breach on his blog, as well as a good deal of eye-opening background on the lead researcher (Emil Kirkegaard) in question. The breach is one not only of informed consent but also the principles of the open data movement.

At the moment, Kirkegaard is unrepentant, and Arhus University (where he is currently a grad student) has issued a tepid response. As for Kirkegaard's "research", it largely speaks for itself. Much of it appears in a particular "journal" that is edited by, you guessed it, Emil Kirkegaard. Over half of the entries published in the stable of "journals" at Openpsych.net on which he is listed as editor (Open Differential Psychology, Open Behavioral Genetics, and Open Quantitative Sociology and Political Science) are ones authored or coauthored by Kirkegaard. In reality, Keyes is spot on when he refers to the journal as little more than a blog. The editorial boards consist of individuals who are either pseudonymous or ones who might not be obvious fits for the subject matter for each journal (I'll go ahead and use the term, albeit loosely), and the peer review process appears to be minimal at best. Although Beall's List is exclusively devoted to predatory publishers that amount to little more than vanity presses charging exorbitant fees to their authors with little to no peer review, at least some subset of what Kirkegaard is engaged in is predatory in a much deeper sense. Beall might wish to consider adding Kirkegaard's journals to the list as a warning to anyone biased toward the open data movement who might be suckered into thinking his journals are even remotely legit.

If one does look at Kirkegaard's Google Scholar or ResearchGate profiles, it shall become readily apparent that the bulk of the citations his work garners are self-citations. There are a handful of authors publishing in peer-review journals citing his work as well, although those appear for now to be a small minority. I certainly have plenty of questions that I would love to ask his advisor regarding the amount of oversight s/he has over Kirkegaard's work - after all, graduate students are expected to receive considerable mentoring, and clearly something has gone off the rails in this particular instance. In addition, I wonder what it says about an academic department and its host institution that essentially is facilitating the work of an alleged student who appears to be more interested in abusing a position of privilege. As for the OKCupid users affected, whatever one might want to say about internet and privacy, it should go without saying that they did not deserve to be Kirkegaard's victims.

I may say more if time permits.

A personal note

My wife broke a hip late February. Since then I have, in addition to attending to taking on many of her daily responsibilities and dealing with the usual problems that crop up when dealing with insurers and physicians, been simply trying to maintain some semblance of my own work-related responsibilities. After all, there are classes to teach, committee meetings to attend, and data sets to analyze and write up. She had some complications post-surgery that we are still dealing with. She is getting better and will undoubtedly be walking at her normal baseline level before all is said and done. The complications - namely visual from a mini-stroke occurring post-surgery - will take a bit longer to sort out. Blogging obviously was going to take a back seat under the circumstances. I'll have some new posts up as time permits.

Wednesday, February 24, 2016

More on Trump and Authoritarianism.

As I was digesting the news earlier this morning, a couple articles stood out. One from NYT examined the tendency for Trump's supporters to be racially/ethnically intolerant (based on available polling data). The article is careful to point out that Trump hasn't actively encouraged racism, but is tapping into and exploiting a political undercurrent that was already there to begin with. That strikes me as sensible enough.

The other article that grabbed my eye was by Matthew MacWilliams, whom I have mentioned before. Once more he lays out his case for the link between authoritarianism and support for Trump. The other variable that seems to act as a predictor is fear of terrorism. One helpful thing about this particular column was that the author goes on to discuss how he has been measuring authoritarianism (itself a contentious issue in social psychology, political psychology, and political science) and provides at least some of the data analyses (at least in graph form) to make his case. Essentially, he uses a four question instrument to measure authoritarianism that largely focuses on parenting that appears to have been used successfully by previous political scientists studying the construct of authoritarianism. It helps to know how the variable is being measured as we evaluate his work. So far, it appears that those who score higher in authoritarian attitudes show higher levels of support for Trump. What is striking about the author's thesis is that the predictive link between fear of terrorism and support for Trump suggests that contrary to conventional wisdom, there may not be a ceiling for Trump support. To the extent that Trump can successfully exploit terrorist attacks occurring in the US (in particular if there is a link to, say, ISIS, as was the apparent case in the recent San Bernardino mass shooting) or in Europe, he may be able to gain traction as the primary season continues.

I'm also not surprised that Trump would periodically show support for the use of torture, as there is a link between authoritarian attitudes and attitudes for torture - something that has been found in some of my research that is currently in press, as well as by others. Anyway, this is fairly interesting work. Trump won't have to really do much to stoke racial or ethnic resentment, as the link between that variable and authoritarianism is well-established (most recently in an article I published last year), but if he continues to use fear-based rhetoric with regard to potential terrorist threats, he could be a formidable opponent and contender for the GOP nomination as the election season progresses. Assuming he wins the nomination, he could be a more formidable opponent for the eventual Democratic nominee than conventional wisdom has suggested as election season ramps up later this fall. The great unknown is whether our "Authoritarian Spring" will last past the early primaries. We'll see.

Sunday, January 17, 2016

One characteristic that appears to distinguish Trump supporters

According to a national survey of 1,800 registered voters, the variable that best predicted support for Donald Trump was authoritarianism. The idea that highly authoritarian individuals tend to be highly obedient and gravitate toward strong or charismatic leaders is nothing new, as the author duly notes. In the context of the current electoral season, though, it should give us something to consider as the first caucuses and primaries are to be held within a matter of weeks. The author suggests that pollsters may be incorrect in assuming that Trump has a ceiling of support within the Republican Party that he has already reached. Rather, the author of the study suggests that there is the distinct possibility that Trump will fare better than pundits have predicted (including one whom I tremendously respect, Nate Silver, of 538 blog).

The only other variable that the author claims was statistically significant was fear of terrorism, which seemed to indicate support for Trump. Other variables, such as income level, race, etc., did not serve as significant predictors, according to the author. The author's narrative does seem to fit with some data I published last year that showed a significant positive link between authoritarianism and racial/ethnic resentment, social conservatism, and intolerance, which may well be what characterize the attitudes of Trump's base of supporters.

Some skepticism is always in order. There has been some question about how much of the variability authoritarianism actually predicts, with some scholars suggesting that other variables, such as social dominance orientation are better predictors of political attitudes and behavior than authoritarianism.

The researcher behind the survey, Matthew MacWilliams, is a doctoral candidate at University of Massachusetts. His research was conducted very recently and as of yet has not gone through the usual vetting process that academic research normally undergoes. I am guessing or at least hoping that other social scientists will have the opportunity to examine the data as well, and that MacWilliams has the opportunity to present his findings in venues other than a political blog. The results are interesting, nonetheless.

Monday, January 11, 2016

Passings: Leonard Berkowitz

Earlier this month, social psychologist Leonard Berkowitz passed away. You can read his obituary here. It is practically impossible to be an aggression researcher and not encounter his influence. Much of his work focused on media violence and cognitive cues on aggressive behavior. He also was well known for his efforts to advance theory and research on the relationship between frustration and aggression. His own theoretical model, the Cognitive Neoassociation Model, shared quite a bit in common with the theoretical model that has guided much of my own research. Berkowitz was a solid methodologist, and one who wrote quite extensively in defense of the lab experiment - a research approach that was at one point under attack (look up the "crisis" in social psychology). When I was initially admitted to the University of Missouri's Social Psychology doctoral program, my summer reading list included work by Berkowitz, and I was strongly encouraged to get a hold of a copy of his then-current textbook on aggression. One of Mizzou's Social Psychology faculty members, Russ Geen, was in fact one of Berkowitz's students, and Berkowitz was an informal mentor to one of my friends and a current collaborator, Brad Bushman (one of Russ Geen's students).

Leonard Berkowitz continued to publish long after he retired, and although I did not know him particularly well, we did share some correspondence periodically with regard to the weapons effect, a phenomenon that he is credited (along with Anthony LePage) with discovering. His work will undoubtedly continue to be influential for many decades to come. Any aggression researcher studying the weapons effect will be citing his famous article that established the phenomenon. Many of us studying media violence or provocation effects will be citing Berkowitz's key research articles or his theoretical work for the foreseeable future. He has left an impressive legacy. That said, he will be tremendously missed in my particular corner of social psychology.