Isolated social networkers

The paper itself cites precisely one source in the sociology literature, despite taking time out in a footnote to explain to its readers exactly what a Pearson correlation coefficient is. One of the co-authors is Neil Johnson, who is meant to be Director of the Oxford Centre for Computational Finance, which is an interdisciplinary centre, so surely it’s his bloody job to make sure that this sort of embarrassment doesn’t happen (Kieran’s post is fourth on a google search for “eurovision voting” so it’s not exactly hard).

What really gets me about this is that this was a joke item on CT, and now it’s apparently been submitted to “Physica A“, which a bit of research reveals is at least a mid-ranking physics journal. If it gets published, then I don’t want to hear anything about Alan Sokal, ever again.

No damn it. I am not going to let this lie. The information commons and the freedom of information, all well and good. But a man’s jokes are his sacrosanct personal property to be defended with fire, brimstone and a cricket bat to the knackers if necessary. Kieran is in Arizona and thus can’t personally defend this one on the field of honour, but I am more than happy to round up a posse. Bloody “hard science” snobs.

(more seriously, a cursory look at the graph Eszter posted shows that the black dots have many fewer interconnections than the grey ones; the physicists involved in social networks research aren’t reading each other’s work either. So it’s not just a matter of “two literatures”; it’s “one literature, plus a gangload of well-meaning dilettantes”.)

It seems like there is an 800 pound white elephant/gorilla mutant chimera in this particular room, which is the relationship between physicists (universally acknowledged as the greatest, smartest, and most important of all scientists) and lowly, pathetic, undeserving social scientists. Clearly, physicists won’t lower themselves to read sociology journals, and physics is clearly way above the heads of sociologists, so the twain never meet (or something). Obviously, as a biologist, I can see how you are all wrong, and evolution is the only possible explanation for all of it!

More seriously, if physicists who got excited about social networks and the lot of it were to admit that their “insights” were really just them figuring out things other people had been studying for years, they might have to admit that maybe, just maybe, social scientists are smart people too. Also, it is a pretty big investment to learn all the lingo, figure out all the basic theories, read some big heavy textbooks, and master an entire domain of science (and no, I didn’t forget to put the “social” on there), only to find out that you haven’t really added much to the discussion. It’s much easier to skim some popular books on a subject, apply some basic math from your own field (say physics or CS), write a paper, and publish it in a place where your reputation gets it in rather than actual merit; in other words, somewhere where people in the field will never see it.

An unfortunate side note for all of us who want the physicists out of our hair. Early in the development of modern biology (molecular biology and genetics, in particular) a bunch of physicists, inspired by Schrodinger’s “What is Life?” (IIRC) decided to pursue bacteriology, nascent molecular biology, and virology. They were smashingly successful (Nobel Prizes and the like), which led the rest of ’em to think that they can pretty much take over any field they want. It’s high time that we put them in their place.

I find the tone of this post (and the comments) ironic considering that while physicists are often guilty of not reading enough sociology literature, some sociologists are guilty of condescension, jealousy and reciprocal arrogance – have the physicists contributed nothing to the science of social networks? And finally, I note also that just as there are few connections from the physics community to sociology, there are equally few going the other direction. Perhaps these communities evaluate the usefulness of ideas in somewhat orthogonal ways? How does whining to each other (socnet is populated mostly by sociologists) about the physicists move science forward? Have you walked across campus to have a conversation with a physicists about social networks?

I’m expecting Cosma Shalizi (a physicist who does this stuff and is scarily well read in the social sciences (and “damn near everything else under the sun”:http://www.cscs.umich.edu/~crshalizi/notebooks/) to weigh in presently). But to play devil’s advocate, much of this work didn’t get much attention in the social sciences before they started weighing in. Some of Barabasi’s work may have been anticipated thirty years ago by Herbert Simon; but Simon’s work on skew functions didn’t get all that much attention among social sciences afik until physicists began to make this issue-area sexy again. And the main direction of research among network sociologists seems to me to have been much more focused on micro-level patterns of exchange within networks than the kinds of topological effects that most physicists have been working on. As well as having a rather different notion of structure. So yes – I do think that physicists have been guilty of stomping in and not paying enough attention to what people have already done – but they’re not the unwitting slaves of defunct sociologists either.

The issue here doesn’t seem to be that physicists are publishing nonsense, just that they are reinventing the wheel

Well a bit of both really. It’s not nonsense to say that you can apply this mathematics to the Eurovision data; you can do cluster analysis of a plate of scouse if you want to. It *is* complete and utter nonsense, though, as Kieran said in the original Eurovision voting study, to claim that there is any social significance or meaning to the results you get from doing so.

_until physicists began to make this issue-area sexy again_

Well this is exactly the issue, right?

_the main direction of research among network sociologists seems to me to have been much more focused on micro-level patterns of exchange within networks than the kinds of topological effects that most physicists have been working on_

I wouldn’t say that was true, really. There’s a branch of social networks that looks at power and exchange in relatively micro-contexts, but there are plenty of more macro-oriented people too. Just take a look at the titles in the “Cambridge Structural Analysis Series”:http://www.cambridge.org:80/uk/series/sSeries.asp?code=SASS&srt=P, for instance.

I am not sure ranting will actually help the situation. I might be an optimistic, but what I see when I look at your network of citation patterns is an inter-connected network not two isolated clusters.
So is the glass half-full or half-empty?

Without addressing the particular substantive issue here, I’d like to speak up in defense of “reinventing the wheel”.

Let’s say someone wants to forge into a new area (for him or her) of scientific research or technological innovation. There are different approaches that can be taken. One is to begin by carefully doing a complete review of the literature or the state-of-the-art, and then continue by building upon the knowledge that has been gained by one’s predecessors. Another is to forget about reviewing the literature, to forget about finding out what anybody else has done, and to attempt to find one’s own way to the desired goals.

The first approach is what most successful scientists or technologists do, and for very good reasons: it avoids wasted effort, allows for incremental progress, and properly acknowledges the contributions of others. But it also comes with a downside: by starting with the work that others have done, one is naturally guided to doing work along the same lines — extending it, but not radically changing its direction.

The second approach is riskier, because one will have to do a lot more to get something new, and because one *is* very likely to “reinvent the wheel”. It does have a real advantage, however: there is a chance that by following one’s own path, one will develop different ways of thinking about the problem than those whose wheels are being reinvented, and that the final result may be innovative in a way that wouldn’t have happened if one followed the first approach. Perhaps the biggest problem with this approach is that it works best for geniuses, and there are a lot more people who think they fall into this category than actually do…

But nonetheless, for the right person or group of people in the right subject area, being unafraid to reinvent the wheel can be the best way to come up with something really new.

“One contributor, however, made the important point that ‘advancing the science’ is the crucial matter, not personal status.”

Personal status is crucial. That’s why people squabble over priority, and it’s why crap gets published by big names while modest but solid work by nobodies is ignored. The noble ideal of building up the body of human knowledge without concern for personal advancement is just another childish fantasy that shatters on first contact with reality. Science works precisely because it has evolved a method for using conflicting desires for personal aggrandisement to separate wheat from chaff. The appropriate response to sloppy or nonexistent literature searches by people moving into a new field is to serve up a big steaming platter of humiliation in the form of a letter to the editor pointing out that the result they have published is so trivial that experts in the field have already published it on the web as a *joke* – mentioning that it shows up as the fourth search result in an obvious Google search is just the cherry on top.

I’m overstating the case here (mostly because getting into all the little details and niceties would take forever), but I hope I’ve made my point. People double check results, perform exhaustive error analyses and literature searches, design rigorous experiments and all the rest in part to avoid the shame of having to retract results. When they fail to do those things the appropriate corrective is to provide them with a disincentive to repeat the error. I urge Kieran (or one of the other interested parties) to write a nice short letter to the editor pointing out that they have published something that adds nothing to the scientific literature. For added spice, throw in a bunch of citations of papers which cover the subject area in greater detail. Double points if the citation list is more than ten times the length of the letter.

As a physicist studying physics, what I find irritating is seeing papers on social networks polluting Phys. Rev. Lett. and other physics journals. Not because I think the study of social networks is useless or uninteresting, but because it’s not physics. A few months ago I couldn’t believe it when the top weekly story in physics (according to the Physics New Update) was a paper on book marketing.

If you’re going to switch subjects, that’s fine, but please switch journals too. I’ll bet the referee pools for the physicists and social scientists don’t overlap much, and if they did, a lot of this reinventing the wheel wouldn’t happen.

There’s also an economics literature on networks that seems to be similarly disconnected. The problem isn’t one of reinventing the wheel, since the economics stuff is mostly game-theoretic, but I suspect opportunities for cross-fertilisation are going begging.

_Rationality and Society_ did a special issue a few years ago on bringing together network theory and game theory – seemed interesting but a little inconclusive if I’m remembering right.

Kieran: You have your answer right there. The Albert & Babarasi paper is available on ArXiv. If Derek Price actually wanted people to read his paper, he would have added it to ArXiv back in 1976.

I do have some comments, as it happens, but since they, ah, grew completely out of control, they’re posted on my blog, rather than here.
http://bactra.org/weblog/347.html

Let’s see, Physica A is a European journal, isn’t it? Not useless at all, just hasn’t built the same reputation as PRL or the Phys Revs or JGR or (insert acronym of most important journal in physics specialty here). On the other hand, most of my experience with European journals is through Annales Geophysicae.

As I understand it, one of the reasons physicists tend to go looking at new fields is funding. For example, there’s much money to be had in molecular biology/biophysics, where I am told by a number of people there is no comprehensive characterization of how biomolecules are organized or behave, not of the sort that would let you predict how a protein would fold or things like that, just lots of ad hoc stuff. We happen to have a large number of people trained in building up theory, so off they go.

As to reinventing the wheel, we probably have less of a fear of doing that, after all we have three different ways of doing quantum mechanics, one of which was done as someone’s PhD thesis (Schrodinger, Lippman-Schwinger, Feynman), three ways of doing classical mecahnics (Newton, Lagrange, Hamilton), two sources of calculus (Newton, Liebnitz), and lots of other stuff that has been duplicated, for various reasons. We tend to worry much more if something cannot be duplicated, cause that implies that maybe the results from the first time around are just crap.

Power laws are not a novel idea in physics. They are so common in physical phenomena that often the first thing physicists do with non-linear data is to see if a power law can adequately describe it (I’m only a grad student and already I’ve had it up to here with power laws).

I’m the founder and one of the organizers of MeshForum (http://www.meshforum.org), which is a conference on Networks dedicated to this very issue – bringing together experts across disciplines. (We also bring together business professionals, government and military experts, and public sector professionals with academics – so lots of networks being connected).

At this year’s MeshForum (Eszter was one of our speakers), we heard from a wide range of academics. There are many other fields studying networks besides Social Scientists and Physicists (though we had both at MeshForum), other fields include transportation experts (operational engineers for example), economists, professors of communication, mathematicians, and computer scientists (not just those who write agent based simulations but the ones who have been working on routing algorithms for decades work with large scale, complex networks on a routine basis).

It may take a while, but we hope that in some small way we can help connect experts across disciplines.

Shannon

I’m neither a physicist nor a social scientist; I’m a software engineer with a mathematical education. I have, however, implemented some social network analysis software, and read some social network literature in the process. I have observed that the social network literature is largely ignorant of basic graph theoretical and mathematical concepts. An example: to do cohesion analysis, the literature presents an algorithm which starts with some random numbers, may or may not converge, and is usually O(MN^3). What it actually computes is simply the eigenvectors of the weighted adjacency matrix. A little bit of SVD made a big difference.

Oops! I spoke too soon about Cosma’s gem, which is summarized right there in the fourth paragraph of his blog post. My apologies.

You might still want to read the whole thing, though ;-)

“The failure to undertake a proper literature search leads researchers to waste time that could be put to more productive use”–it would be hard to disagree with what John Scott says here, at least on the face of it. But as Andrew C. has pointed out, there’s more to this than just doing things “properly”. A quick Web of Knowledge search on “social networks” produces over 3000 hits. Obviously, it would take a lot of work to check through these potentially relevant articles even before you hit the library and start browsing or searching for books. It’s hard enough to keep up with research in your own field, never mind what goes on in others.

When you get to potentially boundary-spanning subject like social networks, you’ll also fall down if you don’t know the terminology other groups use. If I call it a spade and you call it a shovel than I’m never going to find what you’ve written about it using my ‘spade’ search term. Cross-disciplinary institutions and groups can help get round this problem, but only if you have the right mix of people, and only if you have access to such groups.

Going by what comes up top in Google (not that we, as good researchers, would do that) leads to the same problem as going by what you see cited elsewhere: what Robert Merton called the ‘Matthew Effect’ is reproduced, and the same authors (or papers) get cited again. Nevertheless, this type of approach–doing what others do–is what a lot of researchers, lacking the time to do a proper literature search, rely on. Calling for better reviewing is fine, but the reviewers and the people writing the papers are the same folks–it’s “peer” reviewing, after all–and it would be naive to suppose that this will straightforwardly lead to better science.

Those “basic scholarly skills” to which John Scott refers are essential, but we have to keep in mind their, and our, limitations.

The really objectionable thing about the Eurovision paper is not so much the “reinventing the wheel” aspect as the fact that Kieran H was actually correct first time round to present it as a joke paper and call it an “abuse of the data”. These guys have done basically no research into the process by which the dataset was generated or to what the underlying social reality might have been. They’ve not cleaned it up for host-country effects, different national voting schema or anything like that. They’ve just taken the dataset, crunched the numbers, come up with some post-hoc rationalisations of the results of the calculation and called it sociology.

This is the big danger in blundering into a field you don’t understand, not that you’ll reinvent the wheel, but that you’ll make idiotic mistakes by failing to follow research procedures that are there for a reason.

Not to make too much of this, but of course Derek Price was himself a physicist who moved into social science.

“I started off as neutral in all this, neither being a physicist or social network scientist—but I’m afraid all the whining and lack of substance in the arguments has made me end up feeling like standing up for the physicists!”

Yep. Me, too.

“So I think it is a bit unfair to say that everything physicists are doing has already been done and done better by social scientists, who in many cases don’t have the mathematical backgrounds or interest in specific problems that physicists do.”

I’m almost certain that this is somewhat true, though less so all the time. This is an old, general argument, as mentioned above; and I think both sides have legitimate greivances. Physicists are very arrogant generally and specifically regarding the distinction between what some call “hard science” and “soft science” (the latter which they completely disdain). So there’s little doubt that they long have marched in and ignorantly and arrogantly reinvented the wheel…or worse. Even so, it’s also true that most other sciences have lacked the rigor of physics—and as these sciences have matured those intrusive physicists have probably played an important role. Especially as a science moves into a realm where sophisticated mathematical analysis becomes more worthwhile. And there are of course always entrenched people resisting this and whining about it.

Personally, I think a good example of this is d^2’s complaint above against CS people doing agent simulation in economics. From my perspective, economics is paricularly weak in that it is amenable enough to reductive, physics-like mathematical theory that there has been much more success with this approach than there has been with other social sciences—but ultimately this approach will fail, it’s already failing. I personally don’t think that we’ll achieve mature physics-like basic insights into any of the social sciences by any method other than simulation because of complexity. I think we’ll see success first in economics, because in relative terms, it’s “easy” the same way that physics has been “easy”. In my opinion, which I admit isn’t that humble, most of the other sciences (and particularly the social sciences) are so intractably complex that our current theories are really not (relatively) that much more than narratives. That sounds like a swipe, but it’s not.

Has anybody thought to suggest that the old and oft-quoted Santayana “Those who forget history are doomed to repeat it,” saw might be applicable? Was he talking about the doom of cultures and leaders, or the doom of wheelwright researchers as well?

Or us poor benighted putative readers?

If pressed, I would suggest that a 100% fair solution might be to give no scientist, even a social one, any credit whatsoever for anything they write. It seems a senseless waste of ink and pixels and breath to claim credit for work that is essentially a rephrasing of common sense and others’ efforts, and that energy and space might be better spent just reinforcing the point of the utterance, instead of so diligently associating it with one’s self and one’s career.

“Alas, it is a sad fact that our work depends on getting credit where it’s due,” will be somme straw person’s response.

Prove it. No, really, I mean it: prove it. I kid you not.

How important is your reputation with the readers? What makes you think your publications have anything to do with it? Do the people who read your publications not already know who you are, and what you will say, and how you will say it before they have finished the abstract? Are you that literary and cunning, to have hidden the mysterious secret at the end of the work?

Or are you just signaling your peers and competitors? Of the rest of us, who are essentially disinterested, how many will read your work at all?

What is the most important target audience, do you think, of academic publication? I would venture that it is the author’s ego. We could save a deal of time by acknowledging that, though of course we would have to revisit the whole process at length in public to do so.

And, by the way, who was it that wrote Cosma’s gem of a satire? I can’t recall the entire bibliographic entry. I know I had it around here somewhere….

Amused, Tozier

While I agree that arguing about who did what first is pointless, we should remember that this whole discussion *started* with the post by Eszter claiming: ‘Some physicists have come out with a paper on the Eurovision song contest. Kieran… reported similar findings over a year ago. So much for this being “new research”.’
In my limited view of the world, this is a public accusation i.e. that the Physica A paper by the physicists is work that has been done a year ago, and hence challenging the claim of ‘new research’.
I go back to my earlier challenge: (and I quote):
‘..can someone actually show that all (or in fact any) of the results reported in the Physica A paper, have indeed actually appeared elsewhere in the social science network literature? Or indeed on any website posting such as Kieran’s? And I mean exactly the Physica A results not just saying ‘someone else has used a dendogram’. There is no law against the use of a dendogram in other fields. And the Physica A dendogram (which is just one of the introductory results in their Physica A paper) is way more complete (and seemingly correct) than Kieran’s—plus better explained as to how they got it etc.
So thats the challenge—please cite exactly the source, page and which figure/result in Physica A paper has appeared before
In addition, please state which of the Physica A results is wrong.’

You/we cannot make accusations without being able to back it up — it makes the exchange on this website meaningless otherwise. It is also probably illegal!

So if the accusation isn’t completely true, I would have thought the right thing to do is for the accusers to apologize on the website. That is what would happen in a court of law, if this were a libel case!
Then we can all get on with doing far more important things than whinging without proper justification.

Cosma’s post is excellent.

I am no sociologist or physicist either. I have concluded however that in this discussion there is no reinvention of the wheel, the physicists and the sociologists are not talking about the same thing, and know what they know differently.
Eszter pinpoints the sociological position with the “since social phenomena are extremely complex to study” The paper from the Complex Agent- Based Dynamic Networks is talking about complexity. We are referred to Freeman and Barton and “social control” The paper accused of reinventing the wheel discusses “the group “self-assessment process …in the absence of any central controller.
Another major difference is identified by the comment “It is amusing to know that physics is being seriously applied to social networks ignoring the human machine”
The physicists on the other hand say “The complexity in this system emerges.”
One group has a Newtonian view of a special system, the other considers another of the complex adaptive systems, this one a human one.
It is a sad discussion in a way. The Santa Fe mob decided early in the piece not to include sociologists because if they had they would have gotten a lot of social scientists in with no technical background. There would be no common terms on which to build a common language with which they could talk to each other. (Waldrop) Here we have angst about a paper from the Complex Agent- Based Dynamic Networks who want to establish a truly multidisciniplary cluster, ranging from the physical, biological and computational sciences to the social, economic and political sciences And in which a dialogue across disciplines can flourish and in which a shared language can be developed. This discussion seems to prove the need.
Waldrop also expressed the view that it was a gap that may not have been able to be crossed. This would sadly suggest he was right.