Nature Podcast：论文侦探

来源：Nature自然科研

又到了每周一次的 Nature Podcast 时间了！欢迎收听本周由Shamini Bundell  和 Nick Howe 带来的一周科学故事，本期播客片段讨论科学论文侦探Elisabeth Bik的努力。欢迎前往iTunes或你喜欢的其他播客平台下载完整版，随时随地收听一周科研新鲜事。

Host: Shamini Bundell

First up, reporter Adam Levy has made contact with a scientific sleuth.

Interviewer: Adam Levy

Elisabeth Bik spends her days reading academic papers. Well, actually, reading might not be the right word.

Interviewee: Elisabeth Bik

I can indeed scan hundreds of papers sometimes a day because I just look at the images. I’m just like a four-year-old scanning a book for the pictures.

Interviewer: Adam Levy

Elisabeth is a science consultant based in Sunnyvale, California, and she’s staring at pictures for a reason – to look for patterns. She’s hunting for duplicated images, sometimes which are being flipped, rotated or digitally manipulated. These can suggest mistakes or malpractice on the part of researchers, and Elisabeth shares her finds with journals, as well as by posting publicly. Initially just a hobby, Elisabeth became hooked and is now a full-time image sleuth. I gave her a call to find out why and how she does what she does.

Interviewee: Elisabeth Bik

Once you see those patterns, it’s really hard to not see it anymore. You just see the same cell four, five or sometimes ten times in the same photo. Just to see the same thing over and over again in the same photo, it’s sometimes just hilarious to see it.

Interviewer: Adam Levy

What I’ve found when looking at examples of this kind of thing is that when I just see the example I am hopeless, and then when the suspect area is circled or indicated in some way, it just feels so obvious and I can’t believe I didn’t see it in the first place.

Interviewee: Elisabeth Bik

That is, I guess, my experience. But yeah, it is hard to explain because in the beginning I thought well, this is so obvious, and I would send it to a journal editor and they were like, ‘I don’t see it.’ I’m like, ‘Well, it’s obvious, right?’ And they’re like, ‘No, I don’t see it.’ And then I draw these boxes around it and then they’re like, ‘Oh, now I see it.’ So, I realise that maybe not everybody sees it the same way as I do. I need to point that out.

Interviewer: Adam Levy

And now I understand that you’re doing this not primarily as a job, just doing it for the work itself. So, what is your motivation to be going through, as you say, hundreds and hundreds of papers?

Interviewee: Elisabeth Bik

The reason I do that is because if a science paper has some problems, other people might try to replicate that result but they might find that it’s impossible to do, and if that is because the science paper was maybe fabricated or falsified, then that is a really good reason to flag that paper so that other people are warned that there’s a concern about that paper. So, I find that very rewarding. I feel I’m sending out a message that there’s potentially something wrong with papers and helping others to see that as well.

Interviewer: Adam Levy

And I understand that it’s not just standalone papers sometimes that you uncover issues with. You’ve sometimes uncovered problems in groups of papers by the same academic groups or authors.

Interviewee: Elisabeth Bik

I have found several clusters of papers that were all authored by the same research group, so sometimes I might find two or three or ten papers, and one time, I even found 100 papers all by the same author and they all had image problems, So, some of these cases can be quite big.

Interviewer: Adam Levy

Now, you’re not the only person who does this kind of work, but you’re unusual in that you do this quite openly under your own name.

Interviewee: Elisabeth Bik

Yes, I do. Most people who do this work are posting under a pseudonym, and that is because this work can be quite risky. I’m, of course, criticising other people, but a researcher might maybe not like that so much and decide to sue me, so I need to be very careful in how I word my concerns about a paper. So, I’ll just say, ‘These two images look remarkably similar. Can you please explain?’ And hopefully that will keep me safe.

Interviewer: Adam Levy

One thing that has been a source of criticism is that you publish your findings publicly. You go to certain social media to talk about the things you’ve uncovered. Is there a risk that this approach could be counterproductive, potentially causing actual problems further down the line for investigations?

Interviewee: Elisabeth Bik

It might be. So, when I flag papers then that could leave the authors with the opportunity to destroy the evidence. But on the other hand, of the 800-something papers that I’ve reported in 2014 and 2015, only 30% of them have been either corrected or retracted and the rest is just not touched upon, so they’re still out in the open with their duplicated images. So, it’s very frustrating that it seems that journals or institutes are not really acting upon these allegations, so I try to do it the official way but now I’m posting these things online because it’s faster. I feel I’m flagging these things and at least I’ve warned people in a fast way.

Interviewer: Adam Levy

Do you get, I guess, frustrated by how long these things sometimes take to resolve?

Interviewee: Elisabeth Bik

Yes, I do. It is very frustrating when I find these problems, sometimes in seconds, where I see a duplicated background or a duplicated photo, and it seems so obvious to me, and I report these things through the journals and then I just don’t hear anything back for years. And that’s why I take things sometimes to Twitter and say, ‘Guys, this is a problem, you should have looked at it.’ But yeah, some people have called that ‘trial by Twitter’, and I can see that, but sometimes nothing has been done, even though I reported it five years ago. This paper is still out there and everybody thought it was fine.

Interviewer: Adam Levy

Now, I think many people might hear about this and hear about someone who’s literally going through and looking at images on a personal level and think that’s quite an old-fashioned way to do it. We’ve got artificial intelligence and image analysis tools now. Why is this something that a human is still doing?

Interviewee: Elisabeth Bik

Because it’s really hard to computerise this. I don’t know. I like to think of the voice recognition of ten years ago. I would say, ‘Yes,’ and the computer said, ‘No, you said no.’ ‘No, I said yes.’ And image recognition is even more complex. It’s much harder than people think. But having said that, that will happen.

Interviewer: Adam Levy

Do you think as the software improves, you’ll be out of work, you won’t have anything to do?

Interviewee: Elisabeth Bik

No, I think it will take a while before such software is completely up and running, and there’s still going to be the need for a human reviewing the results, similarly to plagiarism checkers that are widely used in scientific publishing, a human is still needed to decide that this was an okay reuse of an image.

Host: Shamini Bundell

That was Elisabeth Bik. To read more about her work, check out the profile of her in this week’s Nature. You’ll also find a news piece detailing a brand-new, cross-industry initiative to try and catch image issues before publication. There’ll be a link to those in the show notes.

Nature Podcast每周为您带来科学世界的全球新闻故事，覆盖众多科研领域，重点讲述Nature期刊上激动人心的研究故事。我们将话筒递给研究背后的科学家，呈现来自Nature记者和编辑的深度分析。在2017年，来自中国的收听和下载超过50万次，居全球第二。

↓↓iPhone用户长按二维码进入iTunes订阅

 ↓↓安卓用户长按二维码进入推荐平台acast订阅

点击“阅读原文”访问Nature官网收听完整版播客