Two Science Gripes

The following things are annoying me this week (aside from the National Express, whose long and pointless correspondance will get a Chatlog entry as soon as it’s complete, assuming it is ever complete):

  1. Bad science journalism and the BBC’s Have Your Say section
  2. People who sell scientific hardware solutions

First, the bad journalism thing. A rather excellent standup comedian called Matt Kirshen (who you won’t have heard of yet but I expect is headed for bigger things) recently directed me to Bad Science, a blog from the Guardian newspaper about how bad most science journalism is, which as you probably know, is something that does bother me. If you don’t believe me, look through the chatlog for The Final Straw or look up the ongoing argument on my Dave Hitt Is A Twat page. In fact, there is one item that I found on Bad Science that I’d already covered here: the Carex handwash adverts.

But the point is that we need someone like Ben Goldacre (who runs Bad Science) to roam the science debates on the BBC’s Have Your Say website and mark down all the stupid arguments, because as it is we can post comments, and recommend comments, but there’s no way to mark a comment as bad, so you end up with a load of highly recommended comments that are terribly compelling to some people but are patent gibberish to scientists or people experienced in whatever other field is relevant to the debate. And of course, once they have a few recommendations, they acquire some kind of apparent authority, as well as rising to the top of the page, and more people start recommending them, whereas rebuttals to them tend to drop off the page unnoticed. The system just doesn’t really work at all as a way of promoting debate or determining public opinion, at least, not informed public opinion.

My other science gripe this week, in case you’re genuinely that stupid that you’ve forgotten, is scientific hardware solutions you can buy, usually for thousands upon thousands of pounds.

I’ve been writing scientific software for a year now, and sometimes I wonder how much I could make doing it freelance. The problem is that as far as I can tell scientific hardware tends to be sold along the same business model as photocopiers, Parker fountain pens, printers, and games consoles: sell something very cheap and then sell ink or games or whatever at a massive markup. More money in the long run. The scientific hardware version of this seems to be that the hardware is sold cheaply but the software to use it or analyse the data is massively overpriced, and usually pretty shockingly poor.

Case in point being the FLIR infrared camera we’re using. I don’t know how much the camera cost, but I’m told that each copy of the analysis software costs £3000. And it’s shockingly bad, and requires we use Windows NT. I realise it’s running on a computer which I think is about 300MHz, but it does take about a minute to plot a line which I’ve managed to plot in Matlab with one command and in a short enough time that you don’t even notice the pause.

But of course, we can’t use my Matlab routine without spending two days converting the data into a format that takes up about 5 times as much disk space, then spending another week copying and converting it, so that FLIR can sell their advanced software that can do this for us – probably at the same achingly slow pace it does everything else – to any institution that doesn’t mind paying another three grand to save time.

And all this strikes me as a pretty stupid way of doing things. Partly because science is generally run on honesty and facts and I would have thought that would suggest selling each item for what it costs to build and distribute plus markup rather than following the strange and obtuse business tricks and tactics employed by people who sell videogames. Partly it’s because this means that a research institution that only uses one computer to analyse the data will pay so much less than an institution that uses many, even if they do the same experiment (or to flip that around into a more relevant concern, it means that to save money people will scrimp on the software – for example the FLIR system we use has a computer permanantly attatched to it, and therefore any time we want to use the analysis software we need to book the IR camera as well. But mostly it seems stupid because the software is so uniformly awful and the idea of paying thousands of pounds for it is pretty galling.

And that’s a problem with the way things seem to be sold in the field. I don’t know if this happens in other parts of science, but in dental research it’s very hard to buy a QLF machine, or an intraoral camera, or an IR camera. Hardly anybody seems to be interested in selling such things. They sell solutions, not products: they won’t sell you a camera, but they’ll happily sell you their latest image acquisition and archiving package, which comes with software and a camera that will often not work with any other software.

And the software they do give you is always highly incompatible, and rubbish at doing anything except following the exact procedure the manufacturers would like. The QLF software has a proprietary analysis algorithm in it, but you can’t just load an image and analyse it because it doesn’t expect anyone to do that. If the data isn’t in its own format it just can’t see it. It will only run with one particular graphics card that costs a thousand pounds apiece. And it won’t export data. Seriously. It can store data in its own impenetrable format, and it can export it as CSV, but only one record at a time. We have an entire study that’s done and uses the software and we’re actually thinking of emplying someone just to go through and get all the numbers out of it, because it’s going to be a mammoth task of repeatedly clicking the same things and none of us fancy doing it ourselves. At one stage we have the software designer round to ask him about this, and aside from accidentally revealing his master password on our projector screen we didn’t learn much. He promised to have a look at the source code and try to work out the file format so we could decode it, but he didn’t seem confident that he’d manage it and never got back to us.

And all of this is ludicrous, because science doesn’t work like that. It’s an evolving field, and what’s standard practice one year might be shown to be wrong the next year and shunned as being a bit backward the year after. It’s just not practical or useful to sell a product that can only be used in one specific way. That’s very limiting. It would seem to me far more sensible to put out a widely compatible product and give out the spec sheet for it so that we could compare them usefully and so that if we decide to do something with it slightly different to what the manufacturers were expecting we can do it directly rather than conning their software package into letting us do it. They’re far more interested in traping you into their product line than they are in helping you do research, and I think that’s a major problem. If the IR analysis software manual just had the file format of their obnoxiously opaque and bizarre SEQ files then we could do the same research five times as quickly.

The other problem, of course, with selling software rather than hardware is that if I buy the hardware, I can write my own software. I could probably go one better and sell that software. And I could undercut the official software because I’m not recouping money I lost selling expensive equipment.

I think scientists are mature enough to accept it if you’re up front about what your product costs. They know what they’re doing and they’ve done it before, and they’re not going to fall for stupid pricing tricks. Last time we looked at a camera the questions of how much the consumables cost and if it would work with our own software were asked before we even asked for a quote.

I think it’s something that really needs to change, because it seems that there’s an awful lot of time and money being wasted doing things this way.