Saturday, 26 September 2015

Memories of Geoff Lilley

On Monday morning I got a phone call telling me that Professor Geoffrey M Lilley had died the previous day at the age of 95. The news, though sad, wasn't entirely unexpected; he'd been quite frail for some time and several planned excursions to visit him earlier in the year had had to be put off as he hadn't been feeling up to it. This obituary, posted by the University of Southampton refers to his 'wit and skill as a raconteur', and when I emailed some of his former friends and associates I wrote "Doubtless lots more will be written about him in due course - everyone who met him came away with a story'.  I thought I ought to write down my 'Geoff stories'while I remember them and here seems as good a place as any to make a start - he takes a while to make his appearance in this one, I'm afraid.

Non-Standard Analysis

 My first paid work (as opposed to study) at the ISVR was on boundary-layer suction. I needed to brush up my knowledge of boudary-layer theory - the fluid dynamics lectures I'd had from the formidable Professor P O A L Davies as a BEng Engineering Acoustics and Vibration student in the late 80s, while fascinating and challenging, hadn't given me as solid a grounding as our current MEng Acoustical Engineering get nowadays.

Around the same time I read The Problems of Mathematics by Ian Stewart and was intrigued by the chapter about Abraham Robinson's Non-Standard Analysis. During my PhD Joe Hammond, my supervisor, had encouraged me to make contact with David Chillingworth in the Maths department and take his course on Advanced Calculus with Applications. Another fascinating course, this one from a pure mathematician (the 'application' turned out to be that if we were to cut out two particular cardboard shapes, glue them together with cork spacers, stand the result on its side and persuade a heavy enough beetle to walk along one of the perpendiculars, the structure would topple over when the beetle crossed a particular curve) and it introduced me to rigourous methods while showing me how little I, an engineer, knew about that whole area. Non-Standard Analysis seemed to offer a way to formalise the way engineers thought about infinitesimals and, Stewart suggested, allowed results to be obtained that would be much harder to derive by standard methods - 'canards' for instance.

The chapter's last section was called Logic for engineers (no offence, eh?) and mentioned some areas of perturbation theory where it had been applied, one of which was boundary layer flow! This was exciting - perhaps I'd stumbled across a skeleton key that would enable me to unlock wonderful new results in boundary layer theory that couldn't be found any other way? I had to find out more, but all the references for the chapter seemed to be mathematical expositions of the method rather than applications, none more so than Robinson's original book on the subject which I flicked through but found very dense after Stewart's gentle introduction. I asked David Chillingworth if he knew who had applied Non-Standard Analysis to boundary layers. He didn't know but asked Ian Stewart. He couldn't remember where he'd got the boundary layer story from. I searched for references to Non-Standard Analysis in the engineering literature and found that Feri Farassat at NASA had been using them for infinitesimal shock thicknesses, which looked promising but the next time I met Feri I asked him about not only did he not know the boundary layer reference he warned me against using it for perturbation problems at all.

I'd run out of leads when I met Geoff in the staff dining room. (In those less crowded days every table had paper napkins arranged alternately white and coloured - it was understood that the coloured ones were absorbent and were for mopping up spills, while the white ones were for sketching graphs and equations.) Geoff asked what I was up to and I told my tale, somehat surprised by the delight he seemed to be taking in it as I really wouldn't have expected him to have much interest in that sort of thing. His broad grin made it clear he knew something I didn't. The story soon unfolded: when he was at Cranfield College of Aeronautics Geoff had been friends with the inventor of Non-Standard Analysis Abraham Robinson who in the 1940s, despite his main field being logic and analysis, had thrown himself into aerodynamic theory as a contribution to the war effort and become a senior lecturer there. Geoff explained that they used have endless arguments about the importance of rigour and that Geoff had teased him that his ideas were all very well but irrelevant to anything he was interested in. So when 'Abie' published his comprehensive book Non-Standard Analysis he took great delight in giving Geoff a copy and telling him "I've even put in a boundary layer example, just for you!". I went back to the Library and there it was hidden away at the back, a derivation of the basic equations in one paragraph. And if it hadn't been for Geoff it wouldn't even have been there at all.

Sunday, 31 May 2015

Passive Resistance

Professor Lorena A Barba is one of my academic heroes; her 12 Steps to Navier-Stokes changed the way I think about CFD, coding and teaching all at once. If you haven't seen it or its successors then finding out about them is more important than anything that follows in this blog post. Anyway, recently she tweeted something I disagreed with:
but I made a fairly bad job of explaining why on twitter - this post is an attempt to do better.

First I have to refer to another of my academic heroes, Professor Geoffrey K Pullum, whose writing has changed the way I think about language and grammar (and who's no slouch at computer science either), for a thorough explanation of what the passive voice is.  It's far from straightforward and GKP rightly contends that the PV is often misidentified, with both Type I and Type II Errors common (not that I'm suggesting LAB makes either).

So what's wrong with the PV? One complaint is that they obscure agency, as in "decisions were made". Obscuring agency is usually bad, though there are times when it can be done out of sensitivity rather than evasiveness and there are times when the actor is unknown or irrelevant. At one time teachers and editors alike insisted that scientific reports be written entirely in the PV because scientific writing was supposed to be dispassionate - all actions were taken to be the author's, but since procedures should be reproducible it shouldn't matter who performed them, they should be described in enough detail that anyone could do them. This produced plenty of dull writing and the anti-PV movement is, at least in part, a reaction to that.

But as GKP shows there are numerous ways to obscure agency without using the PV. Conversely the PV doesn't have to obscure agency:
Gentlemen, the autopsy reveals that Sir Edgar's face was eaten by zombies!
doesn't leave any doubt about who is being accused.

Another complaint about the PV is that by postponing the revelation of the actor it forces readers to hold images in their heads until that revelation makes sense of them. LAB tweeted a typical example:
The first visible-light snapshot of a planet circling another star has been taken by NASA's Hubble Space Telescope.
That sentence is doing quite a lot of work, as is the way with such journalistic announcements, and one could quibble that it's not clear whether the snapshot is the first ever taken of that phenomenon, or just the first that NASA's HSC has managed to take. But I'm just as happy to be told that

  1. There's a snapshot.
  2. It's of a planet.
  3. The planet is circling another star
  4. The snapshot's been taken (rather than, say, sold).
  5. It was taken by the HSC.
in that order as I am to be told that
  1. The HSC has done something.
  2. It's taken something.
  3. The thing was a snapshot.
  4. The snapshot was of a planet.
  5. The planet was circling another star.
by the active recasting:
NASA's Hubble Space Telescope has taken the first visible light snapshot of a planet circling another star.
By the way, the ambiguity  about the firstness has gone, but now there's a possibility that the HSC is circling another star rather than, or as well as, the planet. But accepting that delayed revelation can be bad there are ways to do it without using the PV; see Kingsley Amis's 'gorged-snake construction':
'Looks like we're out of the woods for a bit', laughed the tall dark sun worshipper as he dubbined with his own hands the boots that...
My real concern about the anti-PV position is that it teaches a rule (don't use the PV) instead of a reason (obscuring agency can withhold useful information; delaying revelations can make reading hard work), and as academics we should be dealing in reasons rather than rule-following. Following the reason would probably result in less use of the PV, whereas following the rule would leave all the ways of achieving the same bad results without using the PV unexamined. There's an analogy with coding; my students often write bad code by using try and catch statements to enclose error-prone routines. Rather than impose a rule (don't use try and catch statements) I prefer to teach a reason (if your routine sometimes produces an error you probably aren't sure it's doing the right thing) as I'm sure most others would.

The problem is that because we have such a tradition of prescriptive, rather than descriptive, grammar we put up with rules far more than we would in other spheres, the most notorious example being the so-called split infinitive. Here the rule is "don't" and the reason is "because I say so" with various cod arguments about imagined derivation from Latin retrofitted. It became a self-fulfilling prophecy and generations learnt to do without this occasionally useful construction and to use alternatives, often clumsier, instead. Now they genuinely find SIs awkward and ugly because they've been conditioned to do so*.

Take coding again: if you don't declare all your variables before using them in a statically-compiled language like Fortran you'll get into trouble, so programmers were taught to always begin a program with declarations for all variables and lost marks when they didn't. When dynamically-typed languages like MATLAB and Python became popular the reason no longer applied but plenty of lecturers still insisted on the zombie rule. Indeed I remember being told that we shouldn't use MATLAB for teaching because it let students get away without writing 'proper' programs, i.e. ones that pre-declared variables. Of course there are other reasons to set up or list some variables at the start of a program (and reasons not to teach MATLAB**).

Now I'm starting to see the PV rule go the way of the SI rule, some people frown when they read a certain clause not because the PV has made it unclear or awkward, it's the passive construction itself that bothers them. And that in turn means that their students are avoiding them because their professors don't like them and for no other reason. It diminishes the language when parts of it are walled off, and it diminishes our students if we tell them to avoid things we don't like. Will the same people who winced or tutted at Star Trek's 'to boldly go' soon do the same when I sing 'These words were composed by Spencer the Rover'?

* I once hurriedly wrote something like "this University's internationally leading research..." to the dismay of a colleague. Had he said no more than that it was inelegant I wouldn't have disagreed, but he insisted that it was grammatically incorrect to separate a possessive from the thing possessed by both an adverb and an adjective, and made dark allusions to split infinitives and the poor state of teaching in my homeland. I assumed, since he was not without a sense of humour, that he was parodying the typical grammar buffoon so I composed a response in the same vein starting "my frightfully dear fellow" and finishing "your humbly obedient servant" with as many other examples of the dreaded construction as I could cram in between. It turned out that he was completely serious. I wish I'd offer to donate money to a charity of his choice if he could find any authority for his 'rule' as long as he reciprocated when he found he couldn't. I just hope his enjoyment of Charlie Brown wasn't spoiled too much when Frieda talked about "my naturally curly hair".

** Or, indeed, to not teach it. For what it's worth I invite you to imagine the content of two public information films, the first entitled "How Not To Paint Your House", the second "How To Not Paint Your House". The first would show an enthusiastic but inept person doing the job very badly; the second would show another snoozing in s hammock while the house in question peels and rots.