Spitting out the AI Gobbledegook sandwich: a suggestion for publishers

 

The past couple of years have been momentous for some academic publishers. As documented in a preprint this week, after rapid growth, largely via "special issues" of journals, they have dramatically increased the number of published articles, and at the same time made enormous profits. A recent guest post by Huanzi Zhang, however, showed this has not been without problems. Unscrupulous operators of so-called "papermills" saw an opportunity to boost their own profits by selling authorship slots and then placing fraudulent articles in special issues that were controlled by complicit editors. Gradually, publishers realised they had a problem and started to retract fraudulent articles. To date, Hindawi has retracted over 5000 articles since 2021*.  As described in Huanzi's blogpost, this has made shareholders nervous and dented the profits of parent company Wiley. 

 

There are numerous papermills, and we only know about the less competent ones whose dodgy articles are relatively easy to detect. For a deep dive into papermills in Hindawi journals see this blogpost by the anonymous sleuth Parashorea tomentella.  At least one papermill is the source of a series of articles that follow a template that I have termed the "AI gobbledegook sandwich".  See for instance my comments here on an article that has yet to be retracted. For further examples, search the website PubPeer with the search term "gobbledegook sandwich". 

 

After studying a number of these articles, my impression is that they are created as follows. You start with a genuine article. Most of these look like student projects. The topics are various, but in general they are weak on scientific content. They may be a review of an area, or if data is gathered, it is likely to be some kind of simple survey.  In some cases, reference is made to a public dataset. To create a paper for submission, the following steps are taken:

 

·      The title is changed to include terms that relate to the topic of a special issue, such as "Internet of Things" or "Big data".

·      Phrases are scattered in the Abstract and Introduction mentioning these terms.

·      A technical section is embedded in the middle of the original piece describing the method to be used.  Typically this is full of technical equations. I suspect these are usually correct, in that they use standard formulae from areas such as machine learning, and in some cases can be traced to Wikipedia or another source.  It is not uncommon to see very basic definitions, e.g. formulae for sensitivity and specificity of prediction.

·      A results section is created showing figures that purport to demonstrate how the AI method has been applied to the data. This often reveals that the paper is problematic, as plots are at best unclear and at worst bear no relationship to anything that has gone before.  Labels for figures and axes tend to be vague. A typical claim is that the prediction from the AI model is better than results from other, competing models. It is usually hard to work out what is being predicted from what.

·      The original essay resumes for a Conclusions section, but with a sentence added to say how AI methods have been useful in improving our understanding.

·      An optional additional step is to sprinkle irrelevant citations in the text: we know that papermills collect further income by selling citations, and new papers can act as vehicles for these.

Papermills have got away with this, because the content of these articles is sufficiently technical and complex that the fraud may only be detectable on close reading. Where I am confident there is fraud, I will use the term "Gobbledegook sandwich" in my report on PubPeer, but there are many, many papers where my suspicions are raised but it would take more time than it is worth for me to comb through the article to find compelling evidence.

 

For a papermill, the beauty of the AI gobbledegook sandwich is that you can apply AI methods to almost any topic, and there are so many different algorithms that can be used that there is a potentially infinite number of papers that can be written according to this template.  The ones I have documented include topics ranging from educational methods, hotel management, sports, art, archaeology, Chinese medicine, music, building design, mental health and promotion of Marxist ideology. In none of these papers did the application of AI methods make any sense, and they would not get past a competent editor or reviewers, but once a complicit editor is planted in a journal, they can accept numerous articles. 

 

Recently, Hindawi has ramped up its integrity operations and is employing many more staff to try and shut this particular stable door.  But Hindawi is surely not the only publisher infected by this kind of fraud, and we need a solution that can be used by all journals. My simple suggestion is to focus on prevention rather than cure, by requiring that all articles that report work using AI/ML methods adopt reporting standards that are being developed for machine-learning based science, as described on this website.  This requires computational reproducibility, i.e., data and scripts must be provided so that all results can be reproduced.  This would be a logical impossibility for AI gobbledegook sandwiches.

 

Open science practices were developed with the aim of improving reproducibility and credibility of science, but, as I've argued elsewhere, they could be highly effective in preventing fraud.  Mandating reporting standards could be an important step, which, if accompanied also by open peer review, will make life of the papermillers much harder.

*Source is spreadsheet maintained by the anonymous sleuth Parashorea tomentella

 

N.B. Comments on this blog are moderated, so there may be a delay before they appear. 

So do we need editors?

It’s been an interesting week in world politics, and I’ve been distracting myself by pondering the role of academic editors. The week kicked off with a rejection of a preprint written with co-author Anna Abalkina, who is an expert sleuth who tracks down academic paper mills – organisations that will sell you a fake publication in an academic journal. Our paper describes a paper mill that had placed six papers in the Journal of Community Psychology, a journal which celebrated its 50th anniversary in 2021. We had expected rejection, as we submitted the paper to the Journal of Community Psychology, as a kind of stress test to see whether the editor, Michael B. Blank, actually reads papers that he accepts for the journal. I had started to wonder, because you can read his decision letters on Publons, and they are identical for every article he accepts. I suspected he may be an instance of Editoris Machina, or automaton, one who just delegates editorial work to an underling, waits until reviewer reports converge on a recommendation, and then accepts or rejects accordingly without actually reading the paper. I was wrong, though. He did read our paper, and rejected it with the comment that it was a superficial analysis of six papers. We immediately posted it as a preprint and plan to publish it elsewhere.

Although I was quite amused by all of this, it has a serious side. As we note in our preprint, when paper mills succeed in breaching the defences of a journal, this is not a victimless crime. First, it gives competitive advantage to the authors who paid the paper mill – they do this in order to have a respectable-looking publication that will help their career. I used to think this was a minor benefit, but when you consider that the paper mills can also ensure that the papers they place are heavily cited, you start to realise that authors can edge ahead on conventional indicators of academic prestige, while their more honest peers trail behind. The second set of victims are those who publish in the journal in good faith. Once its reputation is damaged by the evidence that there is no quality control, then all papers appearing in the journal are tainted by association. The third set of victims are busy academics who are trying to read and integrate the literature, who can get tangled up in the weeds as they try to navigate between useful and useless information. And finally, we need to be concerned about the cynicism induces in the general public when they realise that for some authors and editors, the whole business of academic publishing is a game, which is won not by doing good science, but by paying someone to pretend you have done so.

Earlier this week I shared my thoughts on the importance of ensuring that we have some kind of quality control over journal editors. They are, after all, the gatekeepers of science. When I wrote my taxonomy of journal editors back in 2010, I was already concerned at the times I had to deal with editors who were lazy or superficial in their responses to authors. I had not experienced ‘hands off’ editors in the early days of my research career, and I wondered how far this was a systematic change over time, or whether it was related to subject area. In the 1970s and 1980s, I mostly published in journals that dealt with psychology and/or language, and the editors were almost always heavily engaged with the paper, adding their own comments and suggesting how reviewer comments might be addressed. That’s how I understood the job when I myself was an editor. But when I moved to publishing work in journals that were more biological (genetics, neuroscience) things seemed different, and it was not uncommon to find editors who really did nothing more than collate peer reviews.

The next change I experienced was when, as a Wellcome-funded researcher, I started to publish in Wellcome Open Research (WOR), which adopts a very different publication model, based on that initiated by F1000. In this model, there is no academic editor. Instead, the journal employs staff who check that the paper complies with rigorous criteria: the proposed peer reviewers much have a track record of publishing and be clear from conflict of interest. Data and other materials must be openly available so that the work can be reproduced. And the peer review is published. The work is listed on PubMed if and when peer reviewers agree that it meets a quality threshold: otherwise the work remains visible but with status shown as not approved by peer review. 

The F1000/WOR model shows that editors are not needed, but I generally prefer to publish in journals that do have academic editors – provided the editor is engaged and does their job properly. My papers have benefitted from input from a wise and experienced editor on many occasions. In a specialist journal, such an editor will also know who are the best reviewers – those who have the expertise to give a detailed and fair appraisal of the work. However, in the absence of an engaged editor, I prefer the F1000/WOR model, where at least everything is transparent. The worst of all possible worlds is when you have an editor who doesn’t do more than collate peer reviews, but where everything is hidden: the outside world cannot know who the editor was, how decisions were made, who did the reviews, and what they said. Sadly, this latter situation seems to be pretty common, especially in the more biological realms of science. To test my intuitions, I ran a little Twitter poll for different disciplines, asking, for instance: 

 

 

 Results are below

% respondents stating Not Read, Read Superfially, or Read in Depth

 

Such polls of course have to be taken with a pinch of salt, as the respondents are self-selected, and the poll allows only very brief questions with no nuance. It is clear that within any one discipline, there is wide variability in editorial engagement. Nevertheless, I find it a matter of concern that in all areas, some respondents had experienced a journal editor who did not appear to have read the paper they had accepted, and in areas of biomedicine, neuroscience, and genetics, and also in mega journals, this was as high as around 25-33%

So my conclusion is that it is not surprising that we are seeing phenomena like paper mills, because the gatekeepers of the publication process are not doing their job. The solution would be either to change the culture for editors, or, where that is not feasible, to accept that we can do without editors. But if we go down that route, we should move to a model such as F1000 with much greater quality control over reviewers and COI, and much more openness and transparency.

 As usual comments are welcome: if you have trouble getting past comment moderation, please email me.

Research funders need to embrace slow science

Uta Frith courted controversy earlier this year when she published an opinion piece in which she advocated for Slow Science, including the radical suggestion that researchers should be limited in the number of papers they publish each year. This idea has been mooted before, but has never taken root: the famous Chaos in the Brickyard paper by Bernard Forscher dates back to 1963, and David Colquhoun has suggested restricting the number of publications by scientists as a solution more than once on his blog (here and here).

Over the past couple of weeks I've been thinking further about this, because I've been doing some bibliometric searches. This was in part prompted by the need to correct and clarify an analysis I had written up in 2010, about the amount of research on different neurodevelopmental disorders. I had previously noted the remarkable amount of research on autism and ADHD compared to other behaviourally-defined conditions. A check of recent databases shows no slowing in the rate of research. A search for publications with autism or autistic in the title yielded 2251 papers published in 2010; in 2019, this has risen to 6562. We can roughly halve this number if we restrict attention to the Web of Science Core database and search only for articles (not reviews, editorials, conference proceedings etc). This gives 1135 articles published in 2010 and 3075 published in 2019. That's around 8 papers every day for 2019.

We're spending huge amounts to generate these outputs. I looked at NIH Reporter, which provides a simple interface where you can enter search terms to identify grants funded by the National Institutes of Health. For the fiscal year 2018-2019 there were 1709 projects with the keyword 'autism or autistic', with a total spend of $890 million. And of course, NIH is not the only source of research funding.

Within the field of developmental neuropsychology, autism is the most extreme example of research expansion, but if we look at adult disorders, this level of research activity is by no means unique. My searches found that this year there were 6 papers published every day on schizophrenia, 15 per day on depression, and 11 per day on Alzheimer's disease.

These are serious and common conditions and it is right that we fund research into them – if we could improve understanding and reduce their negative impacts, it would make a dramatic difference to many lives. The problem is information overload. Nobody, however nerdy and committed, could possibly keep abreast of the literature. And we're not just getting more information, the information is also more complex. I reckon I could probably understand the majority of papers on autism that were published when I started out in research years ago. That proportion has gone down and down with time, as methods get ever more complex. So we're spending increasing amounts of money to produce more and more research that is less and less comprehensible. Something has to give, and I like the proposal that we should all slow down.

But is it possible? If you want to get your hands on research funding, you need to demonstrate that you're likely to make good use of it. Publication track record provides objective evidence that you can do credible research, so researchers are focused on publishing papers. And they typically have a short time-frame in which to demonstrate productivity.

A key message from Uta's piece is that we need to stop confusing quantity with quality. When this topic has been discussed on social media, I've noted that many ECRs take the view that when you come to apply for grants or jobs, a large number of publications is seen as a good thing, and therefore Slow Science would damage the prospects of ECRs. That is not my experience. It's possible that there are differences in practice between different countries and subject areas, but in the UK the emphasis is much more on quality than quantity of publications, so a strategy of focusing on quality rather than quantity would be advantageous. Indeed, most of our major funders use proposal forms that ask applicants to list their N top publications, rather than a complete CV. This will disadvantage anyone who has sliced their oeuvre into lots of little papers, rather than writing a few substantial pieces. Similarly, in the UK Research Excellence Framework, researchers from an institution are required to submit their outputs, but there is a limited number that can be submitted – a restriction that was introduced many years ago to incentivise a focus on quality rather than quantity.

The same people who are outraged at reducing the number of publications often rail against the stress of working in the current system – and rightly so. After all, at some point in the research cycle, at least one person has to devote serious thought to the design, analysis and write-up. Each of these stages inevitably takes far longer than we anticipate – and then there is time needed to respond to reviewers. The quality and impact of research can be enhanced by pre-registration and making scripts and data open, but extra time needs to be budgeted for this. Indeed, lack of time is a common reason cited for not doing open science. Researchers who feel that to succeed they have to write numerous papers every year are bound to cut corners, and then burn out from stress. It makes far more sense to work slowly and carefully to produce a realistic number of strong papers that have been carefully planned, implemented and written up.

It's clear that science and scientists would benefit if we take things more slowly, but the major barrier is a lack of researcher confidence. Those who allocate funds to research have a vested interest in ensuring they get the best return from their investment – not a tsunami of papers that overwhelm us, but a smaller number of reports of high-quality, replicable and credible findings. If things are to change we need funders to do more to communicate to researchers that they will be evaluated on quality rather than quantity of outputs.

Matlab vs open source: Costs and benefits to scientists and society

An interesting twitter thread came along yesterday, started by this query from Jan Wessel (@wessel_lab):

Quick thread of (honest) questions for the numerous people on here that subscribe to the position that sharing code in MATLAB ($) is bad open-science practice compared to open source languages (e.g., Python). What should I do as a PI that runs a lab whose entire coding structure is based (publicly shared) MATLAB code? Some say I should learn an open-source language and change my lab’s procedures over to it. But how would that work in practice? 

When I resort to blogging, it’s often because someone has raised a question that has captured my interest because it does not have a simple answer. I have made a Twitter moment to store the rest of Jan’s thread and some of the responses to it, as they raise important points which have broad application.

In part, this is an argument about costs and benefits to the individual scientist and the community. Sometimes these can be aligned, but in this case, they is some conflict, because those who can’t afford Matlab would not be able to run Jan’s code. If he were to move to Python, then anyone would be able to do so.

His argument is that he has invested a lot of time in learning Matlab, has a good understanding of how Matlab code works, and feels competent to advise his trainees in it. Furthermore, he works in the field of EEG, where there are whole packages developed to do the complex analysis involved, and Matlab is the default in this field. So moving to another programming language would not only be a big time sink, but would also make him out of step with the rest of the field.

There was a fair bit of division of opinion in the replies. On the one hand, there were those who thought this was a non-issue. It was far more important to share code than to worry about whether it was written in a proprietary language. And indeed, if you are well-enough supported to be doing EEG research, then it’s likely your lab can afford the licensing costs.

I agree with the first premise: just having the code available can be helpful in understanding how an analysis was done, even if you can’t run it. And certainly, most of those in EEG research are using Matlab. However, I’m also aware that for those in resource-limited countries, EEG is a relatively cheap technology for doing cognitive neuroscience, so I guess there will be those who would be able to get EEG equipment, but for whom the Matlab licensing costs are prohibitive.

But the replies emphasised another point: the landscape is continually changing. People have been encouraging me to learn Python, and I’m resisting only because I’m starting to feel too old to learn yet another programming language. But over the years, I’ve had to learn Basic, Matlab and R, as well as some arcane stuff for generating auditory stimuli whose name I can’t even remember. But I’ve looked at Jan’s photo on the web, and he looks pretty young, so he doesn’t have my excuse. So on that basis, I’d agree with those advising to consider making a switch. Not just to be a good open scientist, but in his own interests, which involves keeping up to date. As some on the thread noted, many undergrads are now getting training in Python or R, and sooner or later open source will become the default.

In the replies there were some helpful suggestions from people who were encouraging Jan to move to open source but in the least painful way possible. And there was reassurance that there are huge savings in learning a new language: it’s really not like going back to square one. That’s my experience: in fact, my knowledge of Basic was surprisingly useful when learning Matlab.

So the bottom line seems to be, don’t beat yourself up about it. Posting Matlab code is far better than not posting any code. But be aware that things are changing, and sooner than later, you’ll need to adapt. The time costs of learning a new language may prove trivial in the long term, against the costs of being out of date. But I can state with total confidence that learning Python will not be the end of it: give it a few years and something else will come along.

When I was first embarking on an academic career, I remember looking at the people who were teaching me, who, at the age of around 40, looked very old indeed. And I thought it must be nice for them, because they have worked hard, learned stuff, and now they know it all and can just do research and teach. When I got to 40, I had the awful realisation that the field was changing so fast, that unless I kept learning new stuff, I would get left behind. And it hasn't stopped over the past 25 years!

Open code: not just data and publications

I had a nice exchange on Twitter this week.

Nick Fay had found a tweet I had posted over a year ago, asking for advice on an arcane aspect of statistical analysis:

I'd had some replies, but they hadn’t really helped. In the end, I’d worked out that there was an error in the stats bible written by Rand Wilcox, which was leading me astray. Once I’d overcome that, I managed to get the analysis to work.

It was clear Nick was now having the same problem and going round in exactly the same circles I had experienced.

My initial thought was that I could probably dig out the analysis and reconstruct what I’d done, but my heart sank at the prospect. However, then I had a cheerful thought. I had deposited the analysis scripts for my project on the Open Science Framework, here. I checked, and the script was pretty well annotated, and as a bonus you got a script showing you how to make a nice swarm plot.

This experience comes hard on the heels of another interaction, this time around a paper I’m writing with Paul Thompson on p-curve analysis (latest preprint is here). Here there’s no raw data, just simulations, and it’s been refreshing to interact with reviewers who not only look at the code you have deposited, but also make their own code available.  There’ve been disagreements with the reviewers about aspects of our paper, and it helped enormously that we could examine one another’s code. The nice thing is that if code is available, you get to really understand what someone has done and also learn a great deal about coding.

These two examples illustrate the importance of making code open. It probably didn’t matter much when everyone was doing very simple and straightforward analyses. A t-test or correlation can easily be re-run from any package given a well-annotated dataset. But the trend in science is for analyses to get more and more complicated. I struggle to understand the methods of many current papers in neuroscience and genetics – fields where replication is sorely needed but impossible to achieve if everyone does things differently and only incompletely described. Even in less data-intensive areas such as psycholinguistics, there has been a culture change away from reliance on ANOVAs to much more fancy multilevel modelling approaches.

My experience leads me to recommend sharing of analysis code as well as data: it will help establish reproducibility of your findings, provide a training tool for others, and ensure your work is in a safe haven if you need to revisit it.

Finally, this is a further endorsement of Twitter as an academic tool. Without Twitter I wouldn't have discovered Open Science Framework, or PeerJ, both of which are great for those who want to embrace open science. And my interchange with Nick was not the end of the story. Others chipped in with helpful comments, as you can see below:

P.S. And here is another story of victory for Open Data, just yesterday, from the excellent Ed Yong.

Who's afraid of Open Data

Cartoon by John R. McKiernan, downloaded from: http://whyopenresearch.org/gallery.html

I was at a small conference last year, catching up on gossip over drinks, and somehow the topic moved on to journals, and the pros and cons of publishing in different outlets. I was doing my best to advocate for open access, and to challenge the obsession with journal impact factors. I was getting the usual stuff about how early-career scientists couldn't hope to have a career unless they had papers in Nature and Science, but then the conversation took an interesting turn.

"Anyhow," said eminent Professor X. "One of my postdocs had a really bad experience with a PLOS journal."

Everyone was agog. Nothing better at conference drinks than a new twist on the story of evil reviewer 3.  We waited for him to continue. But the problem was not with the reviewers.

"Yup. She published this paper in PLOS Biology, and of course she signed all their forms. She then gave a talk about the study, and there was this man in the audience, someone from a poky little university that nobody had ever heard of, who started challenging her conclusions. She debated with him, but then, when she gets back she has an email from him asking for her data."

We wait with bated breath for the next revelation.

"Well, she refused of course, but then this despicable person wrote to the journal, and they told her that she had to give it to him! It was in the papers she had signed."

Murmurs of sympathy from those gathered round. Except, of course, me. I just waited for the denouement. What had happened next, I asked.

"She had to give him the data. It was really terrible. I mean, she's just a young researcher starting out."

I was still waiting for the denouement. Except that there was no more. That was it! Being made to give your data to someone was a terrible thing. So, being me, I asked, why was that a problem? Several people looked at me as if I was crazy.

"Well, how would you like it if you had spent years of your life gathering data, data which you might want to analyse further, and some person you have never heard comes out of nowhere demanding to have it?"

"Well, they won't stop you analysing it," I said.

"But they may scoop you and find something interesting in it before you have a chance to publish it!"

I was reminded of all of this at a small meeting that we had in Oxford last week, following up on the publication of a report of a symposium I'd chaired on Reproducibility and Reliability of Biomedical Research. Thanks to funding by the St John's College Research Centre, a small group of us were able to get together to consider ways in which we could take forward some of the ideas in the report for enhancing reproducibility. We covered a number of topics, but the one I want to focus on here is data-sharing.

A move toward making data and analyses open is being promoted in a top-down fashion by several journals, and universities and publishers have been developing platforms to make this possible. But many scientists are resisting this process, and putting forward all kinds of argument against it. I think we have to take such concerns seriously: it is all too easy to mandate new actions for scientists to follow that have unintended consequences and just lead to time-wasting, bureaucracy or perverse incentives. But in this case I don't think the objections withstand scrutiny. Here are the main ones we identified at our meeting:

1.  Lack of time to curate data;  Data are only useful if they are understandable, and documenting a dataset adequately is a non-trivial task;

2.  Personal investment - sense of not wanting to give away data that had taken time and trouble to collect to other researchers who are perceived as freeloaders;

3. Concerns about being scooped before the analysis is complete;

4.  Fear of errors being found in the data;

5.  Ethical concerns about confidentiality of personal data, especially in the context of clinical research;

6.  Possibility that others with a different agenda may misuse the data, e.g. perform selective analysis that misrepresented the findings;

These have partial overlap with points raised by Giorgio Ascoli (2015) when describing NeuroMorpho.Org, an online data-sharing repository for digital reconstructions of neuronal morphology. Despite the great success of the repository, it is still the case that many people fail to respond to requests to share their data, and points 1 and 2 seemed the most common reasons.

As Ascoli noted, however, there are huge benefits to data-sharing, which outweigh the time costs. Shared data can be used for studies that go beyond the scope of the original work, with particular benefits arising when there is pooling of datasets. Some illustrative examples from the field of brain imaging were provided by Thomas Nichols at our meeting (slides here), where a range of initiatives is being developed to facilitate open data. Data-sharing is also beneficial for reproducibility: researchers will check data more carefully when it is to be shared, and even if nobody consults the data, the fact it is available gives confidence in the findings. Shared data can also be invaluable for hands-on training. A nice example comes from Nicole Janz, who teaches a replication workshop in social sciences in Cambridge, where students pick a recently published article in their field and try to obtain the data so they can replicate the analysis and results.

These are mostly benefits to the scientific community, but what about the 'freeloader' argument? Why should others benefit when you have done all the hard work? In fact, when we consider that scientists are usually receiving public money to make scientific discoveries, this line of argument does not appear morally defensible. But in any case, it is not true that the scientists who do the sharing have no benefits. For a start, they will see an increase in citations, as others use their data. And another point, often overlooked, is that uncurated data often become unusable by the original researcher, let alone other scientists, if it is not documented properly and stored on a safe digital site. Like many others, I've had the irritating experience of going back to some old data only to find I can't remember what some of the variable names refer to, or whether I should be focusing on the version called final, finalfinal, or ultimate. I've also had the experience of data being stored on a kind of floppy disk, or coded by a software package that had a brief flowering of life for around 5 years before disappearing completely.

Concerns about being scooped are frequently cited, but are seldom justified. Indeed, if we move to a situation where a dataset is a publication with its own identifier, then the original researcher will get credit every time someone else uses the dataset. And in general, having more than one person doing an analysis is an important safeguard, ensuring that results are truly replicable and not just a consequence of a particular analytic decision (see this article for an illustration of how re-analysis can change conclusions).

The 'fear of errors' argument is, of understandable but not defensible. The way to respond is to say of course there will be errors – there always are. We have to change our culture so that we do not regard it as a source of shame to publish data in which there are errors, but rather as an inevitability that is best dealt with by making the data public so the errors can be tracked down.

Ethical concerns about confidentiality of personal data are a different matter. In some cases, participants in a study have been given explicit reassurances that their data will not be shared: this was standard practice for many years before it was recognised that such blanket restrictions were unhelpful and typically went way beyond what most participants wanted – which was that their identifiable data would not be shared.  With training in sophisticated anonymization procedures, it is usually possible to create a dataset that can be shared safely without any risk to the privacy of personal information; researchers should be anticipating such usage and ensuring that participants are given the option to sign up to it.

Fears about misuse of data can be well-justified when researchers are working on controversial areas where they are subject to concerted attacks by groups with vested interests or ideological objections to their work. There are some instructive examples here and here. Nevertheless, my view is that such threats are best dealt with by making the data totally open. If this is done, any attempt to cherrypick or distort the results will be evident to any reputable scientist who scrutinises the data. This can take time and energy, but ultimately an unscientific attempt to discredit a scientist by alternative analysis will rebound on those who make it.  In that regard, science really is self-correcting. If the data are available, then different analyses may give different results, but a consensus of the competent should emerge in the long run, leaving the valid conclusions stronger than before.

I'd welcome comments from those who have started to use open data and to hear your experiences, good or bad.

P.S. As I was finalising this post, I came across some great tweets from the OpenCon meeting taking place right now in Brussels. Anyone seeking inspiration and guidance for moving to an open science model should follow the #opencon hashtag, which links to materials such as these: Slides from keynote by Erin McKiernan, and resources at http://whyopenresearch.org/about.html.