Analogies

Speaker A: 00:00:00

Hello and welcome to the Cognitive Engineering Podcast produced by me, Fraser McGruer for Aleph Insights. In this series of podcasts we take a look at interesting topics and discuss what we think they tell us about analysis and decision making. I'm here with Nick Hare and Chris Wragg of Aleph Insights and this week we're discussing what makes a good analogy. Chris, start us off on this. What led us to this subject?

Speaker B: 00:00:29

Well there was a recent sort of article which was looking at the British politician Michael Gove who is a Brexiteer but has remained within the cabinet. He hasn't left Theresa May's government in protest over her negotiating stance and it compared him to Tyrion Lannister and I felt this was a particularly poor analogy. Just refresh my memory, which one's Tyrion Lannister? So Tyrion is the son of the dwarf character played by Peter Dinklage and he's the one who is now siding with Khaleesi, has changed sides, abandoned his family and has probably arguably been the most successful Lannister and has negotiated this complex political environment well whereas to my mind Michael Gove hasn't done so. He was involved in that completely botched attempt to become Prime Minister when he stabbed Boris Johnson in the back and then Theresa May sort of emerged from the melee and also he's abandoned his political base. I mean maybe that's where the comparison works but in lots of places

Speaker A: 00:02:02

it falls down as a general analogy. And also, and I don't say this lightly, I mean Tyrion Lannister is actually very likeable and Michael Gove is famous for not being at all likeable. Yes, I mean I love

Speaker C: 00:02:17

Tyrion, he's one of the best characters on the show and you know it seems to me that his success is more about his use of his intellect to anticipate consequences of his actions and his adaptability. So his ability to be very robust when things go wrong and a lot of the other characters you know fail because they don't do that because they stay with their fixed ideas about what they want to achieve. Yeah very much not like Michael Gove really. Who's an ideologue. Yeah I mean but I wouldn't have described, one thing you wouldn't say that Tyrion is a schemer as such. I mean he doesn't deceive people very much. If anything he tries to persuade people through laying facts out you know he's not a deceiver. I mean having said all that I actually have some, I think I'm the only man in Britain who quite likes Michael Gove. I think he's very bright, I used to like him on the moral maze. I've heard he's a very good minister. Yeah I liked him on the moral maze I think most but he's become a popular, I could see why he'd make a good hate figure. What would be a, well

Speaker A: 00:03:16

let's assume for a moment it's a bad, I know we're going to spread the broadness out, what would be a better analogy for Michael Gove either within Game of Thrones or without?

Speaker B: 00:03:25

Yeah well I think Littlefinger is probably a better analogy for him because he's very self-serving, he doesn't, you know he's abandoned sides quite regularly, he's a bit creepy and

Speaker C: 00:03:43

hopefully ultimately he'll have his throat cut. Yeah the thing about Littlefinger is he thinks he's a great schemer but he's actually an inept schemer because everyone can see through him, everyone knows he's a schemer and that doesn't work. I thought also Grimmer Wormtongue might be a good, which one's that? He's in Lord of the Rings. Yeah he's the one who basically

Speaker B: 00:04:09

takes over King Thid and he whispers in his ear and he's possessed. He's Saruman's agent.

Speaker A: 00:04:19

Well it's a bit like they say in politics, particularly in the Conservative party, is the schemers don't make it as Prime Minister. So going back to Michael Portillo, and it's like you were saying with Littlefinger, it's just so transparent, their scheming. So going back to Michael Portillo, Boris Johnson, arguably with Michael Gove,

Speaker C: 00:04:43

these people just don't sort of make it to the top. No I think that's right and so yes anyway, but we don't really want to talk about Michael Gove and Game of Thrones, well we probably do,

Speaker A: 00:04:52

but we're not going to. So yeah let's go back to the original question and then what does, what makes, we've said that's a bad analogy, what makes a good analogy then? Or how can analogies be useful? Well what, first of all what is an analogy? There we go,

Speaker C: 00:05:05

let's start with that. So and I, it is actually, it's something very fundamental to the way we think. I think if we can sort of abstract what the process of analogising is, probably best to start with sort of the formal field of data analysis. So you know in machine learning and so on, you start with structured data and structured data. It's always data with you isn't it? Go on, keep going. Information data, you call it what you want, but the point is that what you have is a set of fields, so you have data points and what data points are, is a kind of conjunction of values to different fields. Wait hold on, he's just made an analogy about data. We've got name Fraser, age 52 or whatever, you know weight 17 stone, you know, a set of data that is conjoined together. Now what that means is that when we look at at you as a data point, we can look at other people who are the same weight and the same age and that perhaps have the same name and the ability to say right, we've got a group of people who all have the same age here and we can try and find relationships between that characteristic and other characteristics is what all data analysis is. That is what data analysis is, is finding relationships between you know sort of dissimilar data points that nevertheless have similar characteristics. That's what analogies are, it's the ability to say okay we've got these two systems or events or you know people or anything which share some sufficient number of features and often that enables you to say well if they share those features right there's a certain probability that they'll share other features and that's really fundamental to the way we reason is to be able to diagnose a totally new situation as being basically similar to some other situation because it shares features with that situation and that enables you to form beliefs about it. You can say oh okay well I mean if he's a man then we can say that he's probably going to be a bit taller than if he was a woman even though we don't know anything about you because we're able to analogize from you to the set of all other men right. And humans are particularly

Speaker B: 00:07:23

good at analogies you know and it works well for our sort of evolutionary setting because you know we don't have, whereas lots of machine learning approaches might involve the investigation of enormous data sets, for people we've only, we have a small amount of personal experience and from that we need to form you know accurate beliefs about the world and so this process of you know analogous thinking is a good way of doing that of saying yes I saw a thing a bit like that once, I think this thing might be a bit like that in these particular regards. So I mean for an

Speaker C: 00:08:03

analogy to work right, so a successful analogy relies on three things. First of all that your assignment of the features is correct so that you are a man, we haven't got that wrong, rather than you assume the data's right, that comes first. Second you assume that your inference process works so that you know that I've correctly discerned that there is a relationship between you know one variable and another and I'm drawing the right inferences from it. And then finally that this relationship is persistent in the right way so you know that this relationship between gender and height is going to be true let's say in this new area that I've looked at or now like it might be that that relationship just yesterday broke down. And so we make these those three things are you know that are necessary for the success of an analogy and indeed for the success of human thought because all thought is essentially an analogy. So what and I think so it's probably worth saying that the only reason we have evolved to think analogously is because by and large the world we live in is stable. So you know it does things don't change. Tigers don't suddenly switch to being you know to being nice and cuddly and friendly one day to the next you know. Rain clouds don't suddenly produce heat instead of rain right. So the fact that the world is more or less stable is why we've evolved to think like this. Well I mean you can't evolve to think if you don't have a stable world to reason about so yeah there we are. We're an animal that sort of

Speaker A: 00:09:40

has this quest and desire for understanding knowledge and for understanding stuff that we

Speaker C: 00:09:44

don't yet understand. Yeah I know but what I'm saying is that we wouldn't be able to do that right. You couldn't have machine learning at all. You can have any learning machine or human or anything else if there weren't shared features across data points. So if every data point was totally unique. Do you remember we did that podcast about Funes the Memorius, the Borges story where you know every single experience to him was a completely unique conjunction of things and Borges said you know that means he's really unable to form abstract thoughts. So I think the potential

Speaker B: 00:10:20

problem with this analogous thinking is you know the false analogies that John Stuart Mill referred to where you know as Nick was saying you know we break those three criteria. So the example he gave was of you know knowing that somebody's sibling is lazy and assuming that they're lazy which you know probably doesn't work as an analogy or you know I think about with hoverflies being black and yellow they have obviously evolved to exploit false analogy you know that black and yellowness is somehow associated with having a sting of some description. And because we love analogies because we're you know that is that is the strength of our form of sort of reasoning. We are also susceptible to being exploited by the use of false analogy. Yeah I mean I think what we're saying is deception is

Speaker C: 00:11:27

it's possible to be wrong but you can be wrong in the right way. I mean you know it might be the case that well if you know that someone's sibling is lazy they have a higher probability of being lazy. So it might be right to say well you know 10% of people are lazy but there's a genetic component and we've discerned that through analysis of siblings and we've found that you know actually if one sibling's lazy the other one might be 50% probable that they're lazy. That's completely okay right. There's nothing wrong with that. The problem is that often you know we're not very good at reasoning like that. We haven't needed to evolve you know quite those sort of fine distinctions and in everyday life it's very tempting and easy to just seize on two or three features that are shared and assume that they're all shared and that's where it really that's where

Speaker A: 00:12:14

it really seems to go wrong. Incidentally as going back to this example of siblings and also the hoverfly I've just as the hoverfly benefits from a false analogy so have I in that case because I'm a young younger of two siblings and when I was at school all my teachers assumed I was going to be really nice and hard working and a good solid chap because that's exactly what my older brother's like turns out you know that's not the case but I benefited from that. Okay well look I mean where where do we go from here I mean one thing one sort of I remember at university one thing we talked about a lot is is certainly when talking about God for example that we have to talk about if we're trying to understand God we have to talk in analogies because if you try and explain what God is you kind of have to explain him in other ways so it's like a river or it's like it's like an ice that melts or it's and actually so what happens in so in my opinion is you end up believing you're able to construct an abstract but in my opinion it's false

Speaker B: 00:13:17

you know so yeah which the parables and you know all of those those biblical stories that are about you know a returning son or a lost sheep or whatever it might be and so you can

Speaker A: 00:13:29

end up in a false belief in something small like a hoverfly might hurt me or in something maybe huge that that changes history and that there's this God that punishes us

Speaker B: 00:13:38

but also that can work well you know you know analogies are very important for for children's learning well for all of our learning right you know when we're trying to understand a concept an abstract concept so there's been research into the extent to which analogies support learning so when you're trying to understand what a mitochondria does if you're told it's the it's the energy supply of the cell you then start that helps you you you import knowledge from another area it's like it's like an inject of of knowledge oh yes okay I understand it's it's a bit like a battery say for example or you know there are kind of other examples like electricity flows like water and so when you're trying to understand these concepts which are really quite difficult that analogy can make them concrete so in the same way as you can you can maybe exploit that to get across an idea of this abstract being that creates the universe and falsely or otherwise you can also help people learn about things that science have proven you know and so so it's got like any you know feature of reasoning it's got pros and cons and it can be it can be exploited

Speaker A: 00:14:59

sure okay I mean we've got about five minutes or so to go where do we go the use of the use of

Speaker C: 00:15:06

analogies is well documented in the discovery of new sort of mathematical concepts certainly the way that you know I mean so you know you take you find that one problem which is quite difficult to think about actually maps on to a to another problem that is easy to think about I think the best example I could think of was Cartesian geometry where you know geometry had always been treated as different from algebra a totally different thing you know and Descartes kind of said well why don't we why don't we describe these lines and circles using algebraic formulae and then suddenly you know a whole new set of techniques that we developed in algebra could now be applied to geometry and geometry sort of brought into just a form of algebra that that's that's brilliant and it's something very satisfying when we discover a similarity between something complex we don't understand that well and some other thing we do understand it's it's it's and and so that the history of you know certainly of maths because it you know with maths generally something either is absolutely 100% analogous or it's not at all it is full of those kinds of very satisfying fits

Speaker B: 00:16:16

between you know existing problems and new problems yeah and I think you know this is being sort of recognised I mean we talked about machine learning there's a there's an approach called structured mapping engine which was developed by a US academic called Professor Ken Forbus who I instantly trusted because the picture of him showed him as a man with a white beard and glasses and I made the analogy that that's what that's what a scientist should should do but this effectively works by rather than combing lots and lots of data for for patterns simply importing patterns from other learned learned areas so it's being it's being recognised in in technology um I think for for me as as you know as we're using analogy to make decisions the the first thing we we have to do is recognise we are making an analogy because that's I think that's where the problems some of the problems come from where where we're not being explicit about the fact that we're making an analogy so we're importing lots of assumptions about something without actually combing through those assumptions and and looking at them and you know politicians can be very adept at this right so you know there's we're all familiar from English literature with the use of analogy you know of simile and metaphor and we see those coming you know their skin was like a you know a lily or whatever it might be but where it's much more subtle is where use of language imports analogy so when you get people talking about swarms of people or you know a plague of migrants coming to the border or whatever it might be and you you get you subtly start instilling in people's minds an analogy to something else by by use of language and that's where we that's where I think this being explicit about the fact that analogy is going on is helpful to the analytical

Speaker C: 00:18:18

process yeah I'm just bringing the good judgment project for a minute here you know one of the one of the key findings there was that thinking in terms of sort of you know fitting stories and fitting examples to some problem you're thinking about is is generally not helpful in terms of forming accurate beliefs so if you just so in other words if you try and say okay well this this situation you know that Saddam Hussein is a bit like Hitler so you know we'll just import the story of world war ii on to invading Iraq and hey presto you know we just have to do the same thing and it'll be great and and and I you and whereas it's it's you know a more structured process where you say okay I mean it's all and analogies all the way down but you you you look at what aspects are similar you try and identify sort of measurable features of something to say okay well this we're thinking about whether or not this election is going to be fair well you know it's an election so but it's in this particular region you know but it's different in other ways to that region because this country is less corrupt and just you know trying to trying to sort of find features that are similar and features that are dissimilar instead of finding an off-the-shelf story that you could just plonk into a particular situation but yeah I mean I say I think you can see a lot of political debates as really being essentially fighting over which analogy is is most appropriate which is not a good way of doing it but it's how it actually works so I mean I was looking up examples of bad political analogies there's there's there's plenty but a lot of one example people talk about is a lot of libertarians and people conservatives in the US sort of small government conservatives liken taxation to slavery and well it's a good I mean I you know it's got it's not it's not bonkers because essentially you have you are forced to give some of your labor to the government to someone else right so so it's not bonkers but the problem is obviously because of our tendency to import too much of an analogy we therefore inclined to say oh well therefore we should treat taxation as exactly in exactly the same way that we treated slavery and stamp it out you know and those same people are presumably

Speaker B: 00:20:36

the people who 200 years previously have been fighting for slavery yeah so I mean you know

Speaker C: 00:20:43

and so the question is well what's a you know is there an is it is actually is taxation a bit more like you know splitting a bill uh in a restaurant what's it more like you know the in and and so depending on which analogy we decide collectively to go for uh you know the the recommendation will be very different of course you know I mean I would always argue well just stop doing that and just say look this these are the facts about what taxation is and these are the consequences of doing things you know differently um so yeah I mean as Chris was saying basically I think what we're saying is look don't just don't just use analogies be aware of how they work and and what you're legitimately able to do with them and what their limitations are okay uh let's wrap up um

Speaker A: 00:21:26

so let's just return it to the beginning and we were talking about analogies and we were talking about um that we think uh that Tyrion Lannister is not a good analogy for Michael Gove let's look

Speaker C: 00:21:35

at ourselves no I know what's going to happen we're all going to say we're Tyrion Lannister that's that's about everyone wants to be Tyrion Lannister because he's clearly the coolest character in the entire show well look let's start off by being really unfair I'm by the way I'm Tyrion Lannister

Speaker A: 00:21:49

let's start off by being really unfair uh which character would our friend Peter Coghill be which

Speaker C: 00:21:54

which one do we think he is ah you know I feel like there's an answer and I just need to work it out

Speaker B: 00:22:00

yeah yeah let's think about that um I mean he's a little bit like I mean he's he's nowhere near as wet as but he's um in some respects like uh Sam the the guy who goes to study Sam Tarly

Speaker C: 00:22:17

Meister yeah yeah I I don't but I was going to say he's got to be a Meister of some kind yeah because he's very into you know technology and stuff yeah so I I'm veering towards a Meister but I don't think he's quite he's not a soppy and no he's not he's not I tell you who I think he's

Speaker A: 00:22:33

like um who's the female knight what's her name oh um oh yeah uh uh Brie of Tarth yeah I think he's a lot like her because he's she's solid she's stalwart she's fundamentally good um you know when Peter's listening to this I'm really bigging him out he's not he's definitely more he's more

Speaker C: 00:22:51

pragmatic than Brianna she's she's very fixated on her mission okay um and uh I think I don't think Peter would like to see himself as as who do you think Peter would like to see himself as I I suspect he'd go down the Meister route okay um yeah I mean like someone like maybe maybe he one day he wants to be like Meister Lewin who I think is the one in um I might have got my Meisters mixed up but I think he's the one who is at Castle Black and is you know a real sort of you know he's blind really good but he's really but he's you know he's he's he's sort of really

Speaker A: 00:23:23

really hot that's Jim Broadbent isn't it no no I know he's the one in uh yeah yeah anyway okay so that's Peter um you're not allowed to choose yourself okay let's talk about Chris what do

Speaker C: 00:23:35

we think Chris would who Chris would be um I think there's an element of the Cersei about Chris Cersei Lannister really yeah I think there's a bit of a sort of uh I think he can be he can be very ruthless um you definitely don't want to get on the wrong side of him um but uh you know so so yeah he will always he knows what the thing to do is that will be successful and he's got no

Speaker A: 00:24:04

qualms about doing that why what's your what's your feeling yeah um I think that's not a bad one um I mean one thing I often think about Chris he's not the most uh emotionally demonstrative of people um and so I'm trying to think who that is because they're actually they tend to be quite

Speaker B: 00:24:24

can I interject to this yeah go on then because I've always I've always associated massively with Stannis Baratheon okay there we go and there was this there was this point there was this point where he's in Castle Black at some point and somebody says uh you know something something less than and and and I said fewer I said and split second later he corrects their grammar and says fewer and at that point I was his man right up until the point at which he he uh immolates his

Speaker C: 00:24:54

own uh his own daughter yeah yeah I mean I because I do I think but I think actually that that that's also uh that works well and I think uh but I don't I feel like Chris would have succeeded where Stannis failed I think Chris would have would have bided his time a bit longer would have would have waited till till the riper moment to strike and would have and would have won yeah okay yeah

Speaker A: 00:25:15

I think we're getting somewhere with this uh so what about Nick what do you reckon Chris uh I think

Speaker B: 00:25:19

probably uh Khaleesi um from from the perspective of he's he's quite he's quite paternalistic yeah uh he um he's prepared she's very maternalistic but anyway yeah okay okay so but quite um quite prepared to uh make assumptions on behalf of the people she's she's ruling okay um and also she is um uh she's able to to she's always at the center of a set of complex decisions where there's no right answer but she's good at making the least least wrong answer she has a ruthless streak but also prepared to be benevolent and understands the requirement to for social interaction and risk respecting people's views even when yeah when they totally disagree and probably the clincher let's

Speaker A: 00:26:14

not forget likes a good rogering from a big hairy barbarian so uh yeah I mean I'm happy to go along

Speaker C: 00:26:20

with that one uh I I initially I was disappointed not to get Tyrion but I'm happy I'm actually happy with Daenerys Targaryen okay yeah yeah I mean she's she's a great character yeah uh and what about me I was gonna say I feel like you're one of those wildlings one of those hairy wildlings kind of refusing to live by by civilized rules uh you know and uh doing things your own way I'm gonna

Speaker B: 00:26:46

go for uh the red viper um prince prince Oberon he gets his head squished by who gets his head squished by the mountain but who uh who's very artistic and flamboyant uh you know um uh does does things through um uh through artistry rather than uh um you know through a mechanistic approach

Speaker A: 00:27:11

that's right uh yeah no I like that I like I like Oberon I like that sort of uh analogy yeah I would

Speaker B: 00:27:17

say for myself um and also I think you are the kind of person who having sort of bested somebody would be standing over them kind of yeah exactly only to have your head crushed like a melon I think

Speaker A: 00:27:29

I I do like the Oberon thing I would accept that the other one I like is and again I know I never remember any of these characters names but is the sidekick to Tyrion and then later to the northern soldier bloke Robson Robson oh Bron Bron of the Blackwater I love that guy yeah yeah that would be always always got a sort of a little wisecrack going yeah um you know can shimmy his way through

Speaker C: 00:27:51

life yeah no but I see you more of a I don't see you as as enough of a follower really to do that

Speaker A: 00:27:57

properly yeah yeah yeah so for better or worse Prince Oberon it is um well that was good I was

Speaker C: 00:28:04

expecting it to be boring I was expecting us to sort of fixate on some but we picked on some good

Speaker A: 00:28:08

characters there that's good yeah very good um all right um we shall wrap up there I very much enjoyed that so um so we'll wrap up there I'm Fraser McGruer um you've been here with Nick and Chris Wragg of Aleph Insights listening to the Cognitive Engineering Podcast until next time goodbye