About Reasoning and Thinking
I should start by explaining that this is not a review of Manktelow's Reasoning and Thinking, more using it as a point of departure to discuss the pragmatics of thought, the computational and ecological constraints on cognition.
I came to read this book because I had been beginning to write about aspects of cognition, but realised that my own knowledge of the psychological facts and theory was very sketchy and partial. Working in HCI one picks up bits and pieces here and there of different disciplines, ending up with knowledge more like a foreigner's view of a country before visiting: the odd well-photographed landmark (the Eiffel Tower or Coliseum) , and caricature image (the British bobby or French onion seller). Of course, when you actually visit the country there is always the joy of seeing the things that never get shown in cinema or travel brochures., but also often too the shock of how accurate some of the caricature images are.
So, one day in the library I wandered to the psychology section and, like the tourist in a strange city wondering where to visit, was outfaced by the shelves of books. But of course, when in a strange city the obvious thing to do is to follow the trodden tourist trail, so selected Manktelow because it was clearly a student textbook lots of copies, all with popular loan stickers let's find out what the psychology students learn. Checking with colleagues since then, I found it is indeed a course book on the third year of the Lancaster psychology degree.
I can see why it is a chosen text: a clear and well-written text, covering a wide range of areas: simple logical inference, probabilistic inference, decision making, etc. Each topic well illustrated with key experiments and insights from the literature. So if you, like my colleagues in the Psychology Department, are looking for a good textbook, or if you, like me, want to get some insight into the discipline, this is an excellent place to start.
However, my immediate feeling on reading this was a level of shock. Just like the stranger in a foreign land ... do these psychologists really think like that! Of course, the analogy breaks down here. This is a text book, as much a history as a guided tour, and I am a privileged reader, having already seen the surfacing of issues, like being shocked reading a book about Victorian child labour. However, part of me also felt I was on a journey into the psyche of a discipline, more like reading an autobiography. I'll explain.
The book opens with chapters about basic logical inference. Experiments including modes ponens:
if A then B
the classic deduction dating back to Aristotle.
All men are mortal.
Socrates is a man.
Therefore Socrates is mortal.
The experimental evidence was fascinating, the range of different effects studied. People consistently do not think like this, except in very controlled circumstances. Very often we do the 'error' of reversing modes ponens:
All train-spotters wear anoraks.
Martin is a wearing an anorak.
Therefore Martin is a train-spotter.
Of course when we come to conjunctions things get even more confused. Some of these things are known to be due to language: one way 'if' vs. 'if and only if', 'or' meaning 'and/or' or 'exactly one of these'. But other confusions are more deeply seated our logical reasoning is fundamentally 'flawed'.
These effects become more pronounced when the logic relates to real life. Strangely enough (?) we find logic easier when it accords with our prior knowledge than when it disagrees (belief bias).
All pop stars are poor.
Robbie Williams is a pop star.
From this what can you conclude:
(a) Robbie Williams is rich
(b) Robbie Williams is poor
(c) Robbie Williams sings
When we move on to probabilistic reasoning and hypothesis testing, things get even worse. Many of you will know the Wason's card test. Even knowing that it is "not what you think" it is surprisingly hard to choose the 'right' cards to turn over.
In fact I recall many years ago skimming Robert Nozick's PhD thesis (the Harvard extreme libertarian) where he looked at numerous types of potential rational decision strategies and found that human decision making corresponded to none of them.
Wason's selection task
You have four cards before you. Each has a letter on one side and a number on the other.
What cards would you need to turn over to verify the following statement:
"every vowel has an even number on the other side"
Only a minority of people get this 'right', most either fail to turn over a necessary card or turn over an unnecessary card. These 'failures' in testing are surprisingly resilient to restatements of the problem (but do change). However, Cosmides produced a version based around detecting cheating. A newspaper stand has an honesty box and the cards record on one side whether passers by took a newspaper or not and on the other whether they paid.
Which cards do you need to turn over to check that everyone who passes is honest, that is to verify the rule:
"if a newspaper is taken then money must be paid".
When faced with this task the majority of people get it right.
So, what lesson should we take away from this: the perverse irrationality of the human race? In fact the thing I took away was the way in which the various psychologists doing the experiments were pictured as amazed or confused at the apparent lack of 'logic' in their subjects.
The truth is that the ways of thinking that these experiments implicitly suggest are the 'right' ways, are both computationally infeasible (you would be locked in unending reasoning to solve the simplest problems) and ecologically invalid (a real creature in the real world would die if it thought that way).
Perhaps again it is a matter of background, in HCI we are used to seeing both the more applied aspects of cognitive psychology and also the reactions against the more rationalist models of cognition: for example in the work on situated action and distributed cognition and in Winograd and Flores critique of over-formalised cognitive models.
However, it seems that it doesn't require a background in alternative cognitive models to question these assumptions, but more everyday common sense. Recently, I was talking about these issues to someone who teaches exactly this kind of material (I'm not sure which textbook she used). She teaches the standard experiments: the various forms of 'logical' test, the standard Wason experiments etc. just as in this book. But she told me that every time she says to her students "and the experimenters expected people to do X" the students all say "but why" and of course she has no answer because she agrees with her students!
Of course I am also drawing my own caricatures here. Manktelow discusses various alternative kinds of reasoning beyond the simply syllogism including Johnson-Laird's mental models, Rips' PSYCOP theory, and Chatter and Oaksford probabilistic information-gain reasoning.
However these alternatives themselves seem quite divorced from what any person would regard as normal thinking. Throughout the text there is always that sense of experimenters being constantly amazed at their subjects' 'irrational' behaviour and, as reader, my own constant sense of amazement at the experimenters' amazement.
Real reasoning has to be consonant both internally with the constraints of our brains as computational systems and externally with the development of our thinking in a physical and social environment.
Attempts in AI and automated theorem proving very quickly hit the limits of computational power. A complete search for solutions to logical problems using syllogistic reasoning leads to rapid a combinatorial explosion. There are so many potential paths of reasoning to consider that even the fastest computers cannot explore them all. Our brains, which in principle do many things in parallel, could not have the ability to solve even relatively small problems in sensible time frames. Even a game of naughts and crosses (tic-tac-toe) has 362,880 possible game plays and for chess a computer exhaustively examining millions of moves a second would be stopped not by the chess clock but by the expanding sun as it gave its last red gasps of nuclear fusion in 4 billion years time.
The Manktelow book is interesting in that in the very last chapter effectively says "by the way ignore all the previous chapters they are really rubbish". At this point he unpacks some of the computational limitations of reasoning including Simon's notions of bounded rationality and satisficing.
Unfortunately. I'd guess this would only get reached at the last lecture of term when all the students that even manage to attend are recovering from end-of-term parties :-)
Looking at other books in the same general area I notice that Garnham and Oakhill's "Thinking and Reasoning" does discuss some of the limits to sensible rational thinking early on, but then it is a much heavier (in both senses of the word) volume.
Surely it would be possible to succinctly say early on in any exposition of human reasoning that there has been a progression in thinking from more formal models of rationality to more natural models?
Now to be fair when I look again Manktelow's introductory chapter does raise some of these issues was I doing the other classic student error of skipping the first lecture as well as the last?
As well as internal computational validity, rationality must be externally valid; it must make sense in the world that we were developed for the world of the hunter gatherer.
It was good to see in the heart of the book reference to the work of evolutionary psychologists in identifying the different ways we reason about social situations and the way in which this radically alters success in the Wason test.
It is clear that we as humans have not developed significantly neurologically or physically in the last few tens of thousands of years, so any models of cognition ought to make sense for a caveman as well as homo technis.
A lovely example of this is the use of mental models for reasoning and the 'surprise' evidenced at belief bias and various forms of leaking of world knowledge into experiments. The 'mental models' from the Johnson-Laird school all are minimal, in the sense that they include only things stated as part of the problem. This is for good reasons, economy of representation given limited working memory. However, in a simple, pre-technical and pre-cultural environment each perception is about the real world and thus adds to previous knowledge of the world. So, it is not reasonable or sensible to assume that knowledge given is treated in isolation as a set of independent propositions or as an independent model of a problem space.
Instead, the thinking of the natural world suggests that new information is added to existing knowledge. Sometimes it will conflict with existing knowledge and may be rejected ("pop stars are poor") or may overturn the existing beliefs. When we draw conclusions from new information we do not do it from the new information alone, but from everything we have experienced before. Even economy of representation leads to the same ends as we know that the units of working memory are 'chunks', which are of course derived from our previous knowledge.
The surprising think is not that we exhibit belief bias, but that we can ever escape it and perform 'pure reasoning' at all!
One of the roots of this more classically rational thinking is in our ability to imagine past and future things, and thus the ability to exhibit at least episodically modal thinking (in that we can start to consider a possible future and then 'backtrack' to the present). However, for this review let's focus more closely on a more 'developed' and particularly human skill.
If nothing else it is clear that humankind are talking creatures. There is much discussion (more talking) about the inter-relationships between community, culture, language and technology that emerged in that socio-linguistic Eden 40,000 years or so ago, and we can never be certain of the exact details or dependencies, but beyond doubt language is a central aspect.
Again it was good to see in Manktelow a discussion of Grice's principle of cooperation for dialogue and the conversational maxims:
In particular the maxims of quantity and relevance suggests that when we are talking with someone the things they tell us should hold a special place or have special salience, and are not simply added to our store of knowledge. This gets close to the mental model based purely on the premises given. However, these maxims also suggest that our interlocutor should take into account our shared knowledge in framing each utterance and so we must assume and bring in that knowledge. So although the explicitly stated things are rightly more salient it would be unreasonable and wrong to expect them to be the only things in our minds.
The thing that, more than the dialogue itself, perhaps forces us to specially mark or recall those things told in this dialogue as opposed to those previously held is the possibility of deceit, although in this respect the difference between deceit, planning of possibilities and simple story-telling is not so great. If the things we are told may not be, or may not yet be, or may only possibly be true, then we need to be able to mentally add then remove them from our minds. But this is a more sophisticated level of reasoning and indeed the Cosmides version of the Wason test shows that these skills are quite reasonably more finely honed in the social domain than the physical.
Turning from model building to logic, dialogue is still central. Of course the root of the word logic is precisely the Greek logos - speech. This gives the game away. Logic is about argumentation, the things we say to one another to convince or explain. It is the result of, but not the processes of thinking. This is true even of the rarefied reasoning of mathematics and no less the mundane reasoning of day-to-day life. Many of the paradoxes of human reasoning become more clear when we properly recognise this distinction.
In summary of Manktelow I learnt a lot from reading this book: a lot about people's reasoning and perhaps even more about psychologists minds!
However, when reading this or other books in the area never forget that it is unreasonable to be reasonable and never leave your common sense behind when thinking about thinking.
Lena Cosmides (1989). The Logic of Social Exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition, 31:187276.
Alan Garnham and Jane Oakhill (1994). Thinking and Reasoning. Blackwell.
H. Paul Grice (1975). Logic and conversation. In P.Cole & J.P. Morgan (ed.), Syntax and semantics, Vol 3: Speech acts. New York, Seminar Press.
Ken Manktelow (2000). Reasoning and Thinking, Psychology Press, Taylor and Francis.
Alan Dix 6/3/2004