In the previous post I suggested that it’s a mistake to try to understand scientific activity (including misconduct and culpable mistakes) by focusing on individual scientists, individual choices, and individual responsibility without also considering the larger community of scientists and the social structures it creates and maintains. That post was where I landed after thinking about what was bugging me about the news coverage and discussions about recent suicide of Yoshiki Sasai, deputy director of the Riken Center for Developmental Biology in Kobe, Japan, and coauthor of retracted papers on STAP cells.

I went toward teasing out the larger, unproductive pattern I saw, on the theory that trying to find a more productive pattern might help scientific communities do better going forward.

But this also means I didn’t say much about my particular response to Sasai’s suicide and the circumstances around it. I’m going to try to do that here, and I’m not going to try to fit every piece of my response into a larger pattern or path forward.

The situation in a nutshell:

Yoshiki Sasai worked with Haruko Obokata at the Riken Center on “stimulus-triggered acquisition of pluripotency”, a method by which exposing normal cells to a stress (like a mild acid) supposedly gave rise to pluripotent stem cells. It’s hard to know how closely they worked together on this; in the papers published on STAP. Obokata was the lead-author and Sasai was a coauthor. It’s worth noting that Obokata was some 20 years younger than Sasai, an up-and-coming researcher. Sasai was a more senior scientist, serving in a leadership position at the Riken Center and as Obokata’s supervisor there.

The papers were published in a high impact journal (Nature) and got quite a lot of attention. But then the findings came into question. Other researchers trying to reproduce the findings that had been reported in the papers couldn’t reproduce them. One of the images in the papers seemed to be a duplicate of another, which was fishy. Nature investigated, Riken investigated, the papers were retracted, Obokata continued to defend the papers and to deny any wrongdoing.

Meanwhile, a Riken investigation committee said “Sasai bore heavy responsibility for not confirming data for the STAP study and for Obokata’s misconduct”. This apparently had a heavy impact on Sasai:

Sasai’s colleagues at Riken said he had been receiving mental counseling since the scandal surrounding papers on STAP, or stimulus-triggered acquisition of pluripotency, cells, which was lead-authored by Obokata, came to light earlier this year.

Kagaya [head of public relations at Riken] added that Sasai was hospitalized for nearly a month in March due to psychological stress related to the scandal, but that he “recovered and had not been hospitalized since.”

Finally, Sasai hanged himself in a Riken stairwell. One of the notes he left, addressed to Obokata, urged her to reproduce the STAP findings.

So, what is my response to all this?

I think it’s good when scientists take their responsibilities seriously, including the responsibility to provide good advice to junior colleagues.

I also think it’s good when scientists can recognize the limits. You can give very, very good advice — and explain with great clarity why it’s good advice — but the person you’re giving it to may still choose to do something else. It can’t be your responsibility to control another autonomous person’s actions.

I think trust is a crucial part of any supervisory or collaborative relationship. I think it’s good to be able to interact with coworkers with the presumption of trust.

I think it’s awful that it’s so hard to tell which people are not worthy of our trust before they’ve taken advantage of our trust to do something bad.

Finding the right balance between being hands-on and giving space is a challenge in the best of supervisory or mentoring relationships.

Bringing an important discovery with the potential to enable lots of research that could ultimately help lots of people to one’s scientific peers — and to the public — must feel amazing. Even if there weren’t a harsh judgment from the scientific community for retraction, I imagine that having to say, “We jumped the gun on the ‘discovery’ we told you about” would not feel good.

The danger of having your research center’s reputation tied to an important discovery is what happens if that discovery doesn’t hold up, whether because of misconduct or mistakes. And either way, this means that lots of hard work that is important in the building of the shared body of scientific knowledge (and lots of people doing that hard work) can become invisible.

Maybe it would be good to value that work on its own merits, independent of whether anyone else judged it important or newsworthy. Maybe we need to rethink the “big discoveries” and “important discoverers” way of thinking about what makes scientific work or a research center good.

Figuring out why something went wrong is important. When the something that went wrong includes people making choices, though, this always seems to come down to assigning blame. I feel like that’s the wrong place to stop.

I feel like investigations of results that don’t hold up, including investigations that turn up misconduct, should grapple with the question of how can we use what we found here to fix what went wrong? Instead of just asking, “Whose fault was this?” why not ask, “How can we address the harm? What can we learn that will help us avoid this problem in the future?”

I think it’s a problem when a particular work environment makes the people in it anxious all the time.

I think it’s a problem when being careful feels like an unacceptable risk because it slows you down. I think it’s a problem when being first feels more important than being sure.

I think it’s a problem when a mistake of judgment feels so big that you can’t imagine a way forward from it. So disastrous that you can’t learn something useful from it. So monumental that it makes you feel like not existing.

I feel like those of us who are still here have a responsibility to pay attention.

We have a responsibility to think about the impacts of the ways science is done, valued, celebrated, on the human beings who are doing science — and not just on the strongest of those human beings, but also on the ones who may be more vulnerable.

We have a responsibility to try to learn something from this.

I don’t think what we should learn is not to trust, but how to be better at balancing trust and accountability.

I don’t think what we should learn is not to take the responsibilities of oversight seriously, but to put them in perspective and to mobilize more people in the community to provide more support in oversight and mentoring.

Can we learn enough to shift away from the Important New Discovery model of how we value scientific contributions? Can we learn enough that cooperation overtakes competition, that building the new knowledge together and making sure it holds up is more important than slapping someone’s name on it? I don’t know.

I do know that, if the pressures of the scientific career landscape are harder to navigate for people with consciences and easier to navigate for people without consciences, it will be a problem for all of us.

Over many years of writing about ethics in the conduct of science, I’ve had occasion to consider many cases of scientific misconduct and misbehavior, instances of honest mistakes and culpable mistakes. Discussions of these cases in the media and among scientists often make them look aberrant, singular, unconnected — the Schön case, the Hauser case, Aetogate, the Sezen-Sames case, the Hwang Woo-Sook case, the Stapel case, the Van Parijs case.* They make the world of science look binary, a set of unproblematically ethical practitioners with a handful of evil interlopers who need only be identified and rooted out.

I don’t think this approach is helpful, either in preventing misconduct, misbehavior, and mistakes, or in mounting a sensible response to the people involved in them.

Indeed, despite the fact that scientific knowledge-building is inherently a cooperative activity, the tendency to focus on individual responsibility can manifest itself in assignment of individual blame on people who “should have known” that another individual was involved in misconduct or culpable mistakes. It seems that something like this view — whether imposed from without or from within — may have been a factor in the recent suicide of Yoshiki Sasai, deputy director of the Riken Center for Developmental Biology in Kobe, Japan, and coauthor of retracted papers on STAP cells.

While there seems to be widespread suspicion that the lead-author of the STAP cell papers, Haruko Obokata, may have engaged in research misconduct of some sort (something Obokata has denied), Sasai was not himself accused of research misconduct. However, in his role as an advisor to Obokata, Sasai was held responsible by Riken’s investigation for not confirming Obokata’s data. Sasai expressed shame over the problems in the retracted papers, and had been hospitalized prior to his suicide in connection to stress over the scandal.

Michael Eisen describes the similarities here to his own father’s suicide as a researcher at NIH caught up in the investigation of fraud committed by a member of his lab:

[A]s the senior scientists involved, both Sasai and my father bore the brunt of the institutional criticism, and both seem to have been far more disturbed by it than the people who actually committed the fraud.

It is impossible to know why they both responded to situations where they apparently did nothing wrong by killing themselves. But it is hard for me not to place at least part of the blame on the way the scientific community responds to scientific misconduct.

This response, Eisen notes, goes beyond rooting out the errors in the scientific record and extends to rooting out all the people connected to the misconduct event, on the assumption that fraud is caused by easily identifiable — and removable — individuals, something that can be cut out precisely like a tumor, leaving the rest of the scientific community free of the cancer. But Eisen doesn’t believe this model of the problem is accurate, and he notes the damage it can do to people like Sasai and like his own father:

Imagine what it must be like to have devoted your life to science, and then to discover that someone in your midst – someone you have some role in supervising – has committed the ultimate scientific sin. That in and of itself must be disturbing enough. Indeed I remember how upset my father was as he was trying to prove that fraud had taken place. But then imagine what it must feel like to all of a sudden become the focal point for scrutiny – to experience your colleagues and your field casting you aside. It must feel like your whole world is collapsing around you, and not everybody has the mental strength to deal with that.

Of course everyone will point out that Sasai was overreacting – just as they did with my father. Neither was accused of anything. But that is bullshit. We DO act like everyone involved in cases of fraud is responsible. We do this because when fraud happens, we want it to be a singularity. We are all so confident this could never happen to us, that it must be that somebody in a position of power was lax – the environment was flawed. It is there in the institutional response. And it is there in the whispers …

Given the horrible incentive structure we have in science today – Haruko Obokata knew that a splashy result would get a Nature paper and make her famous and secure her career if only she got that one result showing that you could create stem cells by dipping normal cells in acid – it is somewhat of a miracle that more people don’t make up results on a routine basis. It is important that we identify, and come down hard, on people who cheat (although I wish this would include the far greater number of people who overhype their results – something that is ultimately more damaging than the small number of people who out and out commit fraud).

But the next time something like this happens, I am begging you to please be careful about how you respond. Recognize that, while invariably fraud involves a failure not just of honesty but of oversight, most of the people involved are honest, decent scientists, and that witch hunts meant to pretend that this kind of thing could not happen to all of us are not just gross and unseemly – they can, and sadly do, often kill.

As I read him, Eisen is doing at least a few things here. He is suggesting that a desire on the part of scientists for fraud to be a singularity — something that happens “over there” at the hands of someone else who is bad — means that they will draw a circle around the fraud and hold everyone on the inside of that circle (and no one outside of it) accountable. He’s also arguing that the inside/outside boundary inappropriately lumps the falsifiers, fabricators, and plagiarists with those who have committed the lesser sin of not providing sufficient oversight. He is pointing out the irony that those who have erred by not providing sufficient oversight tend to carry more guilt than do those they were working with who have lied outright to their scientific peers. And he is suggesting that needed efforts to correct the scientific record and to protect the scientific community from dishonest researchers can have tragic results for people who are arguably less culpable.

Indeed, if we describe Sasai’s failure as a failure of oversight, it suggests that there is some clear benchmark for sufficient oversight in scientific research collaborations. But it can be very hard to recognize that what seemed like a reasonable level of oversight was insufficient until someone who you’re supervising or with whom you’re collaborating is caught in misbehavior or a mistake. (That amount of oversight might well have been sufficient if the person one was supervising chose to behave honestly, for example.) There are limits here. Unless you’re shadowing colleagues 24/7, oversight depends on some baseline level of trust, some presumption that one’s colleagues are behaving honestly rather than dishonestly.

Eisen’s framing of the problem, though, is still largely in terms of the individual responsibility of fraudsters (and over-hypers). This prompts arguments in response about individuals bearing responsibility for their actions and their effects (including the effects of public discussion of those actions and about the individual scientists who are arguably victims of data fabrication and fraud. We are still in the realm of conceiving of fraudsters as “other” rather than recognizing that honest, decent scientists may be only a few bad decisions away from those they cast as monsters.

And we’re still describing the problem in terms of individual circumstances, individual choices, and individual failures.

I think Eisen is actually on the road to pointing out that a focus primarily on the individual level is unhelpful when he points to the problems of the scientific incentive structure. But I think it’s important to explicitly raise the alternate model, that fraud also flows from a collective failure of the scientific community and of the social structures it has built — what is valued, what is rewarded, what is tolerated, what is punished.

Arguably, one of the social structures implicated in scientific fraud is the first across the finish line, first to publish in a high impact journal model of scientific achievement. When being second to a discovery counts for exactly nothing (after lots of time, effort, and other resources have been invested), there is much incentive for haste and corner-cutting, and sometimes even outright fraud. This provides temptations for researchers — and dangers for those providing oversight to ambitious colleagues who may fall prey to such temptations. But while misconduct involves individuals making bad decisions, it happens in the context of a reward structure that exists because of collective choices and behaviors. If the structures that result from those collective choices and behaviors make some kinds of individual choices that are pathological to the shared project (building knowledge) rational choices for the individual to make under the circumstances (because they help the individual secure the reward), the community probably has an interest in examining the structures it has built.

Similarly, there are pathological individual choices (like ignoring or covering up someone else’s misconduct) that seem rational if the social structures built by the scientific community don’t enable a clear path forward within the community for scientists who have erred (whether culpably or honestly). Scientists are human. They get attached to their colleagues and tend to believe them to be capable of learning from their mistakes. Also, they notice that blowing the whistle on misconduct can lead to isolation of the whistleblower, not just the people committing the misconduct. Arguably, these are failures of the community and of the social structures it has built.

We might even go a step further and consider whether insisting on talking about scientific behavior (and misbehavior) solely in terms of individual actions and individual responsibility is part of the problem.

Seeing the scientific enterprise and things that happen in connection with it in terms of heroes and villains and innocent bystanders can seem very natural. Taking this view also makes it look like the most rational choice for scientists to plot their individual courses within the status quo. The rules, the reward structures, are taken almost as if they were carved in granite. How could one person change them? What would be the point of opting out of publishing in the high impact journals, since it would surely only hurt the individual opting out while leaving the system intact? In a competition for individual prestige and credit for knowledge built, what could be the point of pausing to try to learn something from the culpable mistakes committed by other individuals rather than simply removing those other individuals from the competition?

But individual scientists are not working in isolation against a fixed backdrop. Treating their social structures as if they were a fixed backdrop not only obscures that these structures result from collective choices but also prevents scientists from thinking together about other ways the institutional practice of science could be.

Whether some of the alternative arrangements they could create might be better than the status quo — from the point of view of coordinating scientific efforts, improving scientists’ quality of life, or improving the quality of the body of knowledge scientist are building — is surely an empirical question. But just as surely it is an empirical question worth exploring.

______
* It’s worth noticing that failures of safety are also frequently characterized as singular events, as in the Sheri Sangji/Patrick Harran case. As I’ve discussed at length on this blog, there is no reason to imagine the conditions in Harran’s lab that led to Sangji’s death were unique, and there is plenty of reason for the community of academic researchers to try to cultivate a culture of safety rather than individually hoping their own good luck will hold.