The Human-AI Decision-Making Dichotomy

“Machines are amazing, but they can’t do certain things that only human beings can, like exhibit consciousness, ethics, or the ability to develop social bonds involving emotions, trust, loyalty, and empathy.”

– Andrew Likierman

What you will learn

• Understanding the six elements of good judgment
• How intuition and experience shape decision-making
• Balancing gut feel and logical reasoning in choices
• The impact of awareness on better judgment
• Differences between human judgment and AI capabilities
• Why context shifts are crucial in decision-making
• Integrating human and AI for more effective outcomes

Episode Resources

People

• Herbert Simon
• Danny Kahneman
• Malcolm Gladwell
• Karl Wieck
• Tim O’Reilly
• Heraclitus
• University of London
• Harvard Business Review
• AI (Artificial Intelligence)
• Industrial Revolution
• Pattern recognition

Books

Blink: The Power of Thinking Without Thinking by Malcolm Gladwell
Judgement at Work: Making Better Choices by Andrew Likierman

Transcript

Ross Dawson: Andrew, it’s a delight to have you on the show. 

Andrew Likierman: Ross, thank you very much for inviting me.

Ross: So you have had a long and illustrious career with all sorts of interests that you’ve dealt with over time, and you have spent a lot of time now thinking about judgments. How have you come to this point? 

Andrew: Well, look, I’ve had the pleasure and privilege of working in commercial organizations, in public life and in academic life, and what I’ve seen wherever I’ve been is that judgment is a very, very important quality. And I was intrigued a few years ago to think about the question, all right, so what is judgment? How do we know somebody’s got it? How can we improve our own? If it’s so important, then why aren’t we talking more about it? Why aren’t we including it more? So my work has been to try and pin down what judgment is and how we can use it, in the face of many people who’ve said, Oh, it’s all, you know, you can’t possibly do that. You know, it’s sort of out there. We don’t know quite what it is. Well, I believe we do know what it is, and it helps, because we can help them to improve it.

Ross: Well, I think it’s a very important quest, because some people have good judgment, others don’t, and there seems to be very little in really structured ways to be able to help improve that. So in a relatively recent Harvard Business Review article, and I believe your forthcoming book, you’ve laid out a framework for what are the key elements and how it is we can improve those. So can you share that in a nutshell?

Andrew: Of course, look, I won’t go into very much detail, but just in outline. The reason for having a framework is so that we can identify what it is we need to do to exercise good judgment. Because rather than just thinking vaguely, you know, am I exercising good judgment, and was that a good choice? The framework helps to identify the kind of things one ought to be looking at. And just to be completely clear, I’m not suggesting that you go through this in a mechanical way. What I’m suggesting is that identifying any element of this framework is better than nothing, and the more I believe one can go through the framework and adopt what it suggests, the better one’s chances of making a good choice. So what is it? It’s got six elements. The first one starts with what we know and our experience relevant to whatever it is we’re making a choice about. And I’m going to take an example of going on holiday. Let’s say we go to a place which is very familiar to us, and we’ve been there many years already, so we’ve got lots of knowledge and experience. We know what to expect, where the beach is, where the good restaurants are, and so on. If we’ve not been to this place before, it’s all exploration. We can do a lot of work beforehand, but actually we’ve got to make a lot of, often quite difficult choices, because we don’t know. We haven’t got that experience. 

So the first thing in any choice is, what is the relevant knowledge and experience we’ve got? Then we go on to the question of awareness. When we enter any situation, we need to be aware of what’s going on. And again, taking the holiday analogy, if we go into one part of town and we think, ‘oh is this all right? Is this safe? Or is this not safe? I don’t know’. That’s the kind of thing one needs to be aware of, whereas if you’ve been there again many times before, you’re already aware of what’s going on. But that quality of awareness, which gives one the ability to think through what’s going on here, what’s going on with people, what’s going on in the room, and so on. That’s a very important part of judgment. Number three is the question of who and what we trust. And again, taking the holiday analogy, if we go into a new place, we might get some reviews. Now, the question is, are these reviews we can trust, or can’t we trust them? Are there people who are going to make money out of us going there, we can’t really trust them, are they really independent sources, we can trust them. Our ability to trust in the information we get and the people we talk to when we make a choice is the third element of judgment. 

Number four is the feelings and beliefs that we have. We all have feelings and beliefs. We talk a lot about biases. We talk a lot about emotions. Now, these color the way we make our choices. We need to be aware of what these are. So the question of emotion, and again, if we look at holidays, if everybody wants to go to the seaside, then frankly, suggesting that we go and look around some museums won’t go down very well with members of the family. So we know about that. Feelings and beliefs are an important element of our choice. Number five, we come to actually making a choice. So there we’ve got a question, have we weighed up all the alternatives? Do we know what they are? How good are we at weighing up alternatives? All the things we’ve already said go into making the choice, but the process itself is pretty important. Have you actually gone through the options that might be suitable for this? Finally, there’s the question, so what happens as a result? Because if we make a decision as a result of a choice, the question is, can we deliver it? There’s no saying we’d love to go to Botswana, but if we can’t afford to go to Botswana, then that wasn’t a very good choice, was it? So on the holiday front, on this business of the way that we make our choices, actually carrying something through is part of the good choice we make. So there you are. Those are the six elements,

Ross: Lovely. So I want to dig in and find lots of angles on that. Part of there is, there is a kind of a process there. There are steps and stages in which you can be aware of how well you are taking particular approaches. But underlying this is also that I think you’re starting with experience. You need to have relevant experience. So this goes back to my mind, coming back to Herbert Simon’s work in particular, on pattern recognition. The brain is a wonderful pattern recognition device. Part of it is we expose ourselves to sufficient patterns, we develop this unconscious intuition around recognizing those patterns are similar or not. So are there ways in which we can better feed that experience and pattern recognition which leads to good judgment in similar or different circumstances?

Andrew: Yes, there is, and again, on the business, for example, of awareness. We may have an immediate reaction to something which is based on that pattern recognition, but actually what we need to be aware of is, is that relevant in these circumstances? It’s no use saying I’ll always react to something in exactly the same way if we don’t have a sense of the context. We need that sense of context. Why is this the same as or different from anything we’ve done before? It’s no use having an automatic reaction to something, because we’re going to get some things wrong. Because life moves along. Stuff happens. There is change. So the argument here is, you can’t just simply rely on pattern recognition in order to make a good choice. You have to think about whether this is the same or different to what I’ve experienced before and what I’ve done before.

Ross: Which goes to the point that I think many decision makers to whatever degree, have to grapple with this. This is a logical approach, and this is my gut feel. And there’s all sorts of gurus who point to trust in your god or whatever else. But it is a dilemma, in a sense, where I have this feeling, and then I’m trying to use my logical mind to break this down into a different problem. So are there ways in which we can effectively integrate this, these aspects into decisions?

Andrew: I believe there is. Now, as you’ve said, there are many authorities who have worked on this and done a lot of the really important work on the question of the role of intuition. But what struck me as an outsider coming to this field is how wide the range of authorities is in terms of the kinds of conclusions they draw. So you have some people who say, and Danny Kahneman is a notable exponent, as it were, who is deeply distrustful of the role of intuition. And you have others, let’s say Malcolm Gladwell and Blink, who are very, very fond of the idea that actually, it really matters. Now, in between those, there are many, many different sets of people who have worked on this. So I would contend, on the terms of judgment, that what matters here is, have we done something before? Do we have the context? Do we have the experience and the knowledge on this particular thing? So I would argue that if we’ve done something many times before, and we have a deep knowledge of it, then what we call our intuition is actually really valuable, because it’s all that accumulated knowledge and experience which comes immediately to the fore and says, ‘this is a fake. I know it’s a fake because I’ve looked at 5000 fakes already, so I don’t have to think very deeply about it. I just know it is’. Now, that’s called intuition. What’s also called intuition is someone who, faced with something which is completely new, goes with their gut, without any basis for going with their gut. They have no logic to this. They have no experience. It’s just a kind of feeling. Now I would argue that what differentiates these two is risk. 

If you know something very well and you’ve got a lot of experience of it, then bluntly, your intuition probably carries very little risk. If you’ve never done it before, it’s really risky. Frankly, to do something, and I’m sure we’ve all seen that, people who just do something really stupid because their gut feel tells them to do it, but they’ve never had any experience of it, they don’t know the consequences. So I would argue that what matters here is how much experience and knowledge does one have relevant to this. If one does, then I think intuition can be a great guide. If it doesn’t, it’s really risky. And if you want to take a real risk, well go ahead then, but you know, you might just want to wait, consult, think about it even for five seconds before you make something, which is a decision you’re going to regret. If it’s a complex one, you probably need to bring other people in, get other views, and really allow something to take a bit of time. When talked about the question of sleeping on something, if it’s an unfamiliar and highly risky situation, bluntly, sleeping on something’s probably rather a good idea. Look, if you’ve got no option, if a child is running into the street ahead of you, you don’t wait to think, ‘do I know about this? Do I have experience’? You don’t. You try and save the child, that’s straight away. So there you are. There’s my suggestion about how to make sense of intuition, as far as I’m concerned.

Ross: One of the key elements to your framework is awareness, but what strikes me is that, in fact, all of the elements are around awareness, or self awareness, to metacognition, essentially. It is about being aware of one’s biases, being aware of one’s degree of experience, being aware of the degrees of the frameworks which we have. I suppose taking that meta point, are there ways in which we can enhance or develop our self awareness across those domains, in ways that will enhance our judgment? Because I think there’s many where they might read your article and understand these concepts, but they’re not necessarily going to become more self aware.

Andrew: No, and, look, most of us don’t sort of go through a course on awareness, you know? I mean, we don’t think a lot about awareness. We just kind of assume we’ve got it. But what I’m suggesting is we ought to be aware of how aware we are. And there are lots of ways in which you can make yourself more aware. Just giving an example, if it’s in an organization, if one’s involved in some kind of annual appraisal, or, something which is a regular review of what one does. That’s the kind of time where you can pick up whether you’re aware of what’s going on, and actually, perhaps even ask about certain things, where one says, do you feel, ‘actually, that I’m aware enough of certain things’? 

Training. One can go and get trained in this area. One’s observation skills. There’s lots of ways in which one can do it, and that may be part of other training. So for example, if one goes on a course on relationships between groups, dealing with people, awareness is an important part of that. And if one feels ‘actually, I’m not always aware, as I should have been of certain kinds of things’, you can perhaps seek out some courses that give you an insight into your own awareness. Having a coach or a mentor might be a very good idea, somebody who can tell you, are you picking up the signals? Because you can talk about your relationships with your colleagues and so on, and perhaps they will help you on the question of, ‘well, maybe you should be a little bit more aware of this and that’. Just to give a straightforward example: Quite often, when I’m interviewing people, then they do the talking. That’s the whole idea. Sometimes, when I’m being interviewed, the person interviewing me does all the talking. 

Now it strikes me as very interesting, and I don’t mind that. That’s absolutely fine if they want to talk. That’s great, but it seems to me they’re perhaps not as aware they should be of the fact that, yes, they are doing an awful lot of talking, and maybe they should be giving a lecture rather than interviewing me, because that would probably be more sensible. That’s such a trivial example, but lack of awareness even on things, all of us, I’m sure, have got issues with, are we secure when we go online? Are we about to be hacked by somebody and so on. Now, that’s another kind of awareness. And there again, one can get training for something like that, which says these are the things you need to look out for if you want to be aware of people who are up to no good and trying to get your money away from you when you click on this link, you see what I mean? So there’s many different ways one can do it.

Ross: Indeed. As I recall earlier, interviewing Tim O’Reilly on this podcast, he had taken some courses in animal tracking from indigenous native Americans, and found that very useful in terms of his looking for signals in the evolution of the Internet ecosystem.

Andrew: Well, absolutely. And I think it’s something which is actually very interesting, because we all sit, for example, in meetings, and the dynamics of what’s going on in a meeting is really interesting. And I sense often that people don’t pick up the kind of signals of what’s going on in the meeting quickly enough so that they can respond with a point of view to make sure they get what they want done. It’s not just a question of something of a very esoteric quality. It’s very down to earth, being aware of what’s going on around one.

Ross: Yeah, absolutely. The framing you’ve given the article and your work seems to be very much around individuals. So individual cognition essentially for better judgment, but we can also consider organizational or group cognition, echoing, for example, the work of Karl Wieck and I’m interested around how well can this, what seems to be a framing around individual judgment be mapped onto group or organizational cognition or judgment?

Andrew: There’s limits, because obviously groups operate differently to individuals, and one can’t make the read across an exact kind of way in terms of what I’m doing. But there is a parallel, for example, in the way in which groups themselves operate. And I’ve done quite a lot in discussing with people about how you get better value from groups in terms of the way they operate and the way in which collective judgments are made. Now there’s a whole set of literatures on the question of how to get groups to operate better, but as far as I’m concerned, this matters very much on those aspects where the group comes together to make a choice, and that collective choice, the question here is, would it be better to have a group or an individual making a choice? I think that is a real issue. And for people, for example, running organizations or parts of organizations, it may be, for example, that they are convening a group when actually a group is not necessarily the right thing to do, or on the other hand, and perhaps more often, they make a choice themselves, where actually they would have been better to bring a group together. Here, though, one knows what the advantages and disadvantages of groups are. If you’ve got a group where you’ve got a dominant individual and that individual carries everything,  then maybe the group is not very functional. If, on the other hand, you’ve got a group that operates really well, and you bring together many diverse views, then that can be a great way of making a better collective judgment. So this hinges very much on the effective operation of groups. And for example then, the key role of the chair in making sure that all the voices are heard, that the composition of the group is a good composition for this particular choice and so on. I’ve been on many boards in my time, and I’ve known just how well a group can function and how badly it can function. And so perhaps the awareness of that is in itself really, really important, particularly if you have the choice about whether you convene a group or not and how that group operates.

Ross: So, you recently wrote an article in the frame of AI and how AI is using decisions and essentially making the case that for complex decisions that human judgment is vastly superior. Where are the boundaries? Of course, there are many decisions, some of them quite prosaic or domain specific, where machine learning algorithms, for example, can exceed human performance. What are the domains or defining characteristics of the decisions where human judgment is dramatically superior?

Andrew: You mentioned the fact I’d I said that, as it were, human beings were superior to machines. No, I didn’t say that. What I said was that human beings are different to machines and that’s a quite important distinction because what I’m arguing is that though machines are amazing, artificial intelligence is astonishing in what it can do. One cannot fail to be impressed by this. But what I’m arguing is that there is a distinction between a human being and a machine. So machines can do a lot of things much better than human beings. And if I just take the area of medicine, what we know is that machines are very, very good at certain kinds of analysis in terms of looking at, ‘have you got a problem with a mole or something on your skin’? To be quite blunt, if you have, then you’d be wise to get a machine looking at it, not a human being, because although wonderful human beings are, machines are fantastic at this. 

On the other hand, if you come in with two or three things that are wrong with you, including that you’re not feeling very good about life, a machine is not going to pick that up. So what we’re saying here is that machines are very good at certain kinds of things that you can program. You can train the data really well, you’ve got good data quality, and you can interpret it well, this is fantastic. This is what machines are amazing at. What machines cannot do, are certain things that human beings only can do. And so what I wrote was, and forgive me if I give you a bit of a laundry list here, let me go through it. I’m covering a lot of ground here. 

Okay, so I argue that machines don’t have consciousness, intentionality, a sense of context, meaning, conscience, ethics, self belief, through aspiration or ambition, an ability to develop social bonds involving feeling and emotion, trust, loyalty and empathy. Machines may be terrific at lots of things, but they don’t have that. There’s some more. They can’t anticipate: spontaneity, idiosyncrasy, fallibility and contextual shifts. That’s the thing they can’t anticipate. They’re great at doing stuff, but that’s really tough, and they’re not good at, although, you know they are good in some context, at random interactions. We’re talking random here, fluidity and nuance. Now, this is something AI is pretty good at in some contexts, but not in others. But they can’t think abstractly. I mean by definition, they cannot think abstractly. They can do amazing things, and what they can’t cope with is ambiguity and incompleteness, including the relationship between correlation and causation. Okay, now that’s a big laundry list there. You can see why I’m not suggesting that machines can’t do amazing things, but this is a bit the province of the human being. That’s what makes us different, and that’s why, of course, I believe that judgment is going to be increasingly important in an age of AI, because, actually, that’s what human beings are going to do. They are not going to check proofs, they are not going to make cars. They are not going to, you know, do all sorts of stuff that AI is going to do much better than us, but those are the things that are left for us to do,

Ross: Yes. We think very similarly about that. So we’re just taking one of those points you mentioned, which is around context shifts. I think this is a really interesting one. And part of the thing is that, of course, AI is trained on past data, and the world is changing. So it is not able to map its past, what data it has on the past, on how the world is changing, which is the role of leadership, and arguably, the world is moving faster and faster, yet in the same way humans you talked before about judgment being based on relevant experience. However, humans are also in a shifting context where they need to be able to draw an experience which may not be directly applicable to an emerging and shifting landscape.

Andrew: Well, absolutely. Even among human beings, there are clear differences in this way. Some people get a context shift, they go into a different situation and they behave differently, and some people don’t, they go into a different situation, behave exactly the same as they behaved yesterday. They don’t understand that the context has changed. I think it’s Heraclitus, who said it several thousand years ago, which is, no man goes into the same river twice, because it’s not the same man and it’s not the same river. The world has changed between now and the next time we do something. It may not change dramatically, may not change virtually at all. But for a lot of the businesses where we make difficult choices, that notion of ‘is this context the same’ is really important. If you just take it, you hire a consultant, the consultant comes in and gives you an answer which is clearly the same as the answer they’ve given to the last few clients. It has not much to do with what you do or what you want. Now that’s what I mean by a failure to understand context shift. A machine finds it really difficult to understand context shifts, because context shifts are unbelievably subtle. That’s why so often you have to be in the room in order to understand how it shifted.

Ross: Which takes us to the point of human AI complementarity, and you mentioned, said, they’re different. Absolutely. They’re both, we would hope, our complementary capabilities. So what is the path to marrying or integrating or putting together human and AI judgment, let’s call it, or capabilities or decisions in a way which will give us the best possible outcomes?

Andrew: I’m a natural optimist, and so I mentioned feelings and beliefs, and I say one ought to declare one’s feelings and beliefs. So here I am being optimistic. I mentioned the question of medicine, and potentially the role of machines to do a lot of mechanical things, leaving the doctor free to talk to the patient. In exactly the same way I believe that, well, not exactly the same way, I mentioned change of context. I better take my own advice here. So in much the same way, I believe that machines can do a lot of things that human beings are currently doing much better than them, and that leaves the human beings free to do much more value added. Things that give satisfaction, things that actually provide the element of judgment. And in my own field, for example, of teaching, if I think about teaching at university, what matters here, the interaction between the human beings is really the very valuable bit of this. 

You can get a lot of stuff bluntly to be done mechanically and online. The rote teaching does not have to be somebody standing up at the front of a classroom and talking to other people. That is pretty inefficient and not very satisfying. If the students can do a number of these things online, if they can have ways in which they are helped to learn, leaving them the interaction with the teacher, one which is question and answer and gives value added, it seems to me, everybody benefits. So there’s another example. In a legal firm, it might be that the machine does all the proof checking and does a lot of the grunt work, as it’s called,  that juniors do. The machine can do all that, leaving the juniors actually to get the benefit of the advice from the seniors and to work with them.

Ross: So, role allocations, as opposed to integrated workflows, or just taking that into a for example, a board decision making, board complex decision the role of human judgment is clear in terms of being able to make high stakes human impact decisions in complex and highly ambiguous situations. But within that context, are there specific ways that AI can support or assist or complement or lead to better decisions than humans alone?

Andrew: Well, again, yes. If I think, just as an example, a lot of financial information is produced in a quiet mechanical way by the finance function. It’s based on churning some data. Usually the data is to claim data that was produced last year updated a bit, and it’s produced in a form, bluntly, that’s not very interesting. Now, just applying AI to that alone can give one benefit, one can think about the way in which the planning model has been constructed. It can be a much more sophisticated result of what AI can offer. AI can produce documentation that’s bluntly, more interesting than a lot of humans can, you can get lots of wonderful, whizzy stuff, as a basis of AI. Even in that one single area there, you’ve got a lot of stuff. The marketing side, similarly, on the technology side, there’s a lot that AI can do to help us in our understanding of what’s going on. 

This, of course, means that the individuals around the table have to get to know more about AI and what it can offer. And so that’s a kind of responsibility for everybody to do that. There’s lots of different ways in which, as it were, can come together for boards in that way.

Ross: So just to round out, you already gave us an optimistic vision for where things can go. What is the role of humans in what could be a dramatically different world of work in the coming years?

Andrew: Well, continuing my optimistic theme, it is that human beings potentially have the opportunity to do more of the things that are actually interesting and enjoyable and challenging and less mechanical, as it were, in terms of their roles, and to focus on the areas where they do have comparative advantage, where they can exercise judgment, and that is something which actually adds much more value to an organization and much more value to an individual, actually. That doesn’t mean to say that, you know, there won’t be issues, because, just as in the industrial revolution, hand loom weavers lost their jobs as mechanical looms came in. This is not going to be a seamless transition, and therefore it’s very important that education provide the basis for people to use the technology, to be comfortable with the technology and so on. Because otherwise, just as in the industrial revolution, you’ll get left behind. I think this is a challenge and a real one for people in terms of being able to make sure that society as a whole can cope with the changes involved,

Ross: Absolutely. So is there a title and a release date for your forthcoming book?

Andrew: Well, thank you for allowing me to advertise it. It’s called ‘judgment at work, making better choices’, and it’s coming out on the 23rd of January.

Ross : Oh, fabulous, right. Is there anywhere else where people should go to find out more about your work or the book?

Andrew: If they want to read something before that, as you mentioned, the Harvard Business Review article that I wrote that came out in January, February, 2020, that’s the kind of basis. But I’ve done lots of applications of that. The book contains a lot of applications. If you’re not very impatient, perhaps you could wait until January. 

Ross: Well, I’ll provide links in the show notes to all of those and more of your work. Thank you so much for your time and your insight. Andrew, greatly appreciate it.

Andrew: Thanks very much indeed.