Below, you can find the full transcription corresponding to this episode*:
Julieta Rabanos
Welcome to this new episode of the Heavily Accented Philosophy of Law podcast, powered by the ALF Project. In this episode we have Professor Giovanni Tuzet from Bocconi University. Giovanni, welcome to this podcast.
Giovanni Tuzet
Thanks. Thanks for having me. And hello everyone.
Bojan Spaić
Giovanni, welcome.
Julieta Rabanos
Here we are with Bojan Spaić, as in the other episodes. I am Julieta Rabanos, and as in the other episodes, maybe we can start with some personal questions – or maybe academic personal questions – about our guest. So, Giovanni, can you explain to us what was your academic “way” to the present? Maybe where did you study, what did you study, etc.
Giovanni Tuzet
Yeah, I studied Law in Ferrara, Italy. And I graduated from Ferrara University. Then I did a PhD from the University of Turin and the University of Paris – Paris 12, at the time. The doctoral dissertation was on Charles Peirce: the American philosopher, the pragmatist. On the Italian side, I was doing work in philosophy of law; on the French side, I was doing work in epistemology. Thus, being a philosopher basically started from these philosophical writings and trying to apply some of these ideas to legal reasoning. That was the thing in my dissertation.
Then, after I got the PhD, I spent some time in Switzerland – I was in Lausanne, in Switzerland – working basically in the field of philosophy of science, on abduction (we’ll talk later about abduction, maybe), inference to the best explanation, and models of scientific reasoning.
After that, I came back to Italy and started working at Bocconi University as a legal philosopher. And I’m still there – I’m still at Bocconi.
You want to know also personal stuff about me? I don’t know – you asked me about my academic life, but if you want…
Bojan Spaić
We’ll ask you randomly during the podcast. Just random personal questions.
Giovanni Tuzet
Not too personal, I hope.
Bojan Spaić
We will see.
Julieta Rabanos
We also hope that.
Bojan Spaić
You did also some work on economic analysis of law – in fact, quite a lot of it. How did you stumble upon economic analysis to law, and how did it relate to the things that you were doing regarding Peirce, evidential reasoning, and epistemology?
Giovanni Tuzet
I started doing work on economic analysis of law – or Law and Economics, as some people prefer to call it – when I moved to Bocconi, in fact. It was because of the context, I mean: as everybody knows, Bocconi is a place in which the economic studies are particularly important.
And for this reason, I thought that it was good to do some research and some work in that direction, especially because some people work in the field of legal or on the economic analysis of law or Law and Economics do it without paying too much attention to philosophical issues that are involved in the plans that very often people in the field make. So, I said to myself, maybe there’s need of some work on the foundations, on the philosophical assumptions, on the basic concepts of the economic analysis of law: concepts of efficiency, for instance, the concept of preference or the preferences, concepts like, you know, well-being, wealth maximization, efficiency, wealth, etc.
There are many issues that are philosophical implications and, without disregarding, I mean, the technical work of economists and lawyers that do the details, that work on the details of that, I thought, well, maybe there’s need – I mean, there’s need to have more philosophical works, as I said, and so I try to contribute to the discussion in that respect.
Bojan Spaić
Fantastic. Thanks, Giovanni. Also, you published some recently in the last five years, I think, Giovanni, two very important volumes both on economic analysis of law and on evidential reasoning. We’ll put the links in the show notes when we do the shows.
But let us start with our main topic of today. We’re discussing legal reasoning and its relation to evidential reasoning, or the other way around, so evidential reasoning and its relation to legal.
So, to begin, could you maybe formulate for us an answer to the question or an account of what legal reasoning is?
Giovanni Tuzet
Legal reasoning, very generally speaking, is reasoning about law, legal materials or only legal issues. That doesn’t say a lot, I know it, but the phrase is very, very general. I mean, legal reasoning means a lot of things. So, we need to be more specific about it. It’s a phrase that we use, of course.
I’m not saying that we shouldn’t use the phrase, but it’s a very broad idea. But let me add on my background, on my philosophical background, that I’m basically an analytic philosopher of law.
So, I want to draw the right distinctions, or the distinctions that we need in certain context in order to make names that are specific and testable, etc. So, if you ask me about Legalism in general, my answer would be the one that I just gave you, but more specifically, one has to consider that it’s reasoning that is performed by different subjects.
You have the reasoning of lawmakers, of legislators, for instance, when they discuss about the reasons in favour or against a certain bill that is going to be adopted or enacted. And that’s one thing, which is, by the way, close to political reason. That’s the reason that lawmakers and legislators in particular – it’s not very different from political reasons, if you…
Then there’s reasoning performed by other subjects, actors like, for instance, in law, administrative agencies that perform specific kinds of reason with law.
There’s reasoning performed by lawyers, by scholars, on the one hand, discussing legal issues, commenting on legal cases, materials, etc. And there’s reasoning by lawyers practicing law. So, when you are an attorney and you argue your case for the judge, you perform a piece of reasoning that has some distinctive features, given that you are playing a specific role in the legal practice.
And there’s also – which is the thing I’m more interested in – the reasoning performed by judges. Judicial reason would be the more appropriate term here to use, usually performed by judges when they justify their opinions in those legal systems in which judges have to justify their opinions once they have made the decision.
They write the reasons, I mean, in written form. You have the opinion of the judge or the reasons for certain decisions, and you can follow it and you can see what the reasons are in favour of a certain conclusion, why the arguments of the, let’s say, of the plaintiff were accepted, why the arguments of the defendant were not accepted, what the judges contributed to the case, to the arguing of the litigated points, etc.
That’s the most specific thing that I work on in the field of legal reasoning: judicial opinions, judicial reasoning, or judicial arguments, if you want to call it that.
Bojan Spaić
Maybe now, since you declared yourself as an analytic philosopher, we’ll give you some time to get rid of the Continental philosophers coming with pitchforks and torches in front of your house.
So, this is the – thank you for the concise answer, and we’ll mostly be focusing on, in fact, judicial reasoning and judicial argumentation. But before we pass to that, then I’ll give the floor to Jules.
Could you comment something on this public aspect of reasoning, which is the thing that you mentioned that you will be focusing on, and its relation to this private aspect of reasoning which is or can be a whole other matter?
Giovanni Tuzet
Yeah, that’s again an analytic point. Is there any difference between reasoning and argumentation? Sometimes we use the two words as synonyms, which is not, I mean, a mortal sin, but more specifically one can say that in reasoning you don’t necessarily have the public and controversial aspect that you have in argumentation.
So, when you argue in favour of something or against something, you do it publicly, that is, in front of someone. You have an audience that you want to convince, or you have some interlocutors that you want to convince, to persuade.
That is, in a way, a public dimension that reasoning does not necessarily have, because you can reason on your own and just think about a problem and try to solve it, try to perform your train of reasoning without presenting it to anyone else. When you argue, you argue in front of someone else, right?
So, there is a public instance, dimension, and there’s also a litigated point or a controversial issue, something that needs to be discussed. You don’t argue about things that everybody accepts. I mean, I need not argue on the fact that, I don’t know, Milan is north of Rome, because everyone knows that and everyone agrees on that. But on controversial issues you would have arguments and counterarguments, and in judicial context we need a decision at the end. That’s the business of legal augmentation into initial contexts.
Julieta Rabanos
Talking about this last thing and going back to something – or maybe like thinking of something that you are going to touch when we talk about especially evidence in law and evidential reasoning – how do you connect these public aspects of reasoning in the sense of argumentation with what you have been calling inferentialism in legal reasoning?
Giovanni Tuzet
You know, inferential is legal reasoning is an account of the meaning, the content of claims, to put it like that. So, when you make a claim, the content of it can be made explicit in terms of inferences that support it or inferences that you can draw from the claim.
And that’s the basic intuition of inferentially. So meaning is not something that can be isolated, you know, as the meaning of a sentence. It’s something inferential; it’s something that has to do with inferences that you can perform, or you have to perform, within certain argumentative and linguistic practice. That’s the thing.
And you can apply that to legal matters, to the meaning of legal clauses, provisions enacted by legislators, clauses of contracts, or any kind of legal text that needs to be interpreted. Right, you can have interpretation according to inferential list lines, so interpretation that follows the inferences that are involved in the use of a certain word or the use of a certain concept.
And you can do the same when the claims are about evidence and the facts that the evidence is about. So, claims about testimonies, documents, expert evidence and whatever counts as evidence in law. That’s the very, very basic intuition, of course. It’s much more complicated than this, but to give you the basic idea, I mean, I hope this suffices, right?
Bojan Spaić
Absolutely. So, what would you say, to pass from these general ideas or general accounts of legal reasoning to evidential reasoning, what would you say, then, that evidential reasoning is? Or another way to put it: what could be the nature of evidential reasoning and what would be its role, its kind of position in the legal reasoning?
Giovanni Tuzet
Yeah. Let me use a model that you know, that people who work in this field know quite well, that’s called the syllogism model. That’s a way of capturing what judges should do according to the model, that is, we should draw a conclusion about a case by inferring that conclusion from two premises, essentially: a premise about the law that needs to be applied to the specific case, and the facts of the case.
So, you need a legal norm – can be a rule, typically – and you need the reconstruction of the specific facts of the case to which you are going to apply the law, right. So you have, according to the Synergist model, a major premise stating the law that you need to apply to the case, and you need the minor premise that concerns the facts of the case, right, and then you can draw your conclusion by deduction.
That’s a model that has been criticized in some respects, but we need not discuss those aspects. So, if you buy that idea, you have a minor premise, which is the premise concerning the facts, and the facts need to be reconstructed out of the evidence that the parties present in a case. The parties present the so-called fact finders, judges or jurors, with evidence that is needed to reconstruct what happened or what the specific features of the case are, from a factual viewpoint, I mean.
This is the part of the reasoning, the part of legal argumentation that has to do with the facts, and you need evidence to do that. So evidential reasoning, to answer your question, is the reasoning that has to do with that part of judicial reasoning or judicial communication: the reconstruction of the facts, and in the model, it’s the formulation of the minor premise of the judicial syllogism. That is needed; I mean, you cannot do without it. Otherwise, I mean, what are you talking about in a case if you don’t have the facts of the case, if you don’t establish what happened, if you don’t establish what has been proven or not? I mean, you cannot apply the law, right?
Julieta Rabanos
Mm-hmm.
Bojan Spaić
In isolation from these other parts of, if we take this logistic picture to be broad and kind of giving us a general framework to discuss these issues, in isolation from these other parts, the role of the fact finders in terms of determining the minor premise or the premise about facts seems a lot like factual reasoning of any kind. For example, scientific reasoning, what we do in other disciplines, let’s say in medicine, maybe in science, experimental or empirical research in physics, and so on.
Would you say that, and how does this evidential reasoning law in isolation compare to these other kinds of domains of factual reasoning?
Giovanni Tuzet
There are similarities and there are differences, and the similarities are the ones you are probably thinking of, I mean. When you reason with evidence, you do basically the same thing, whether it’s scientific evidence, whether it’s historical evidence, or whether it’s another kind of evidence, because the basic idea is that you use information and materials that will help you reconstruct some facts that you need to reconstruct, right, in science, in history, in law.
So, there are some similarities between these different efforts in different contexts. But there are also differences. One of the main differences is that in law you cannot just apply any kind of information that is useful, because there are limits in law, and we call them admissibility criteria – that’s a technical thing. So, in law, before a judge, before a jury, you can use evidence only if it passes some admissibility test. We can discuss it later.
So, in the law you have admissibility criteria. You have a test for that. And there are reasons that we have, we are complying with such criteria and tests. You don’t have them in science, in political science, with some exceptions. I mean, we know that some scientific research that has ethical implications should be conducted in a certain way, so there are some limits also.
But basically, there are much less limits than we have in law. History is another example. You don’t have specific limitations in history. You don’t have admissibility criteria, and a historian doesn’t need to say, look, that’s my reconstruction of the relevant facts and look, I have complied with the admissibility criteria – that doesn’t make sense. You just need to have reliable sources and draw reliable and justified inferences from the evidence that you collect: the documents, testimonies, whatever you need to collect in history, right.
In law, there are admissibility criteria, and we can discuss it if you want right now, I mean.
Bojan Spaić
Yeah, our topic that is coming up is actually evidence and evidence law. But Jules, join in.
Julieta Rabanos
But before going to the admissibility criteria, I would have maybe two little questions in regard to what you were saying.
The first one would be, would you say that in law, which is what would be like fact reasoning in law, there is also another limit in the sense of the things that you can consider while you are trying to reason, in the sense of, for example, I was thinking about the fact finders within a trial: that they have only to consider certain assertions of certain means of proof that they have been given, in some sense, which is not a limitation that you necessarily have, I don’t know, in science, that you can basically take within your reasoning everything, not just what other people have given you, and then you’re limited with that.
Giovanni Tuzet
Exactly. The thing you were referring to, I think, is a phenomenon which is a bit different from the one I was referring to earlier. When you have admissibility criteria, you have what young Americans call rules of exclusion, that is, rules that establish what can be admitted and what is to be excluded from a trial.
And here you have the admissibility criteria I was telling you earlier. Next, there’s another phenomenon that sometimes occurs, that is that, for instance, let’s imagine that the testimony of the witness is admitted, but for some reason, at some point the witness makes some assertions that cannot be used in the reasoning to reach a decision, because there is again a rule of exclusion that bans that from the reasoning that should lead to a decision at the end.
And the problem there is that if the assertion, let’s say, was performed in front of the fact finders, now they have it. I mean, they have the information, they know what the witness said, but at the same time they know that, given the rules of exclusion, they shouldn’t use that information or that assertion in their reasoning process.
So, it’s an interesting phenomenon because you should reason as if you didn’t know something that you thought you knew, because you heard the witness saying something. Or, for instance, you are presented with a document that was submitted, but eventually the judge realizes that that document shouldn’t have been admitted. Now the jury have seen the document, and the judge is going to say, look, given that that was inadmissible, now you should do as if you were ignorant about that, which is an interesting cognitive phenomenon, because you should reason in a way counterfactually, imagining that you didn’t know what you in fact know.
So that’s something that has to do with the admissibility problem that I was referring to earlier. You don’t have that in science, you don’t have that in science, and you can put it…
Bojan Spaić
Do you think it would be fair to say that in science we have some… that research in science could be conducted by using some norms, primarily, so norms regulating how truth is to be achieved or how research is to be conducted within science, and eventually we have some informal social norms, and then in law, regarding fact finding, regarding evidential reasoning, we have a wider array of both epistemic, social and legal norms that regulate the fact finding process?
Julieta Rabanos
OK, so coming back to exclusion. We interrupted you when you were going to talk about admissibility criteria. Because we have, from what I can naturally think about evidence in law and then evidential reasoning, we have admissibility criteria, then evaluation criteria, then standards of proof. But maybe we can start with the admissibility criteria that we interrupted you when you were trying to explain.
Giovanni Tuzet
Basically, I said what I wanted to say. I mean, there are many reasons to have admissibility criteria. Some of them are just epistemic, for instance.
Some are economic, because you don’t admit evidence if you have the impression that it’s going to be too costly, given the goods at stake or the stakes of the case, if that will delay excessively the proceedings – economic reasons.
You don’t admit evidence, of course, if it is irrelevant – and that’s the epistemic reason. You don’t admit evidence if it was coerced or illegally obtained. You don’t admit the evidence for a number of reasons. And of course, any legal system has its own criteria or rules of exclusion. Let me add that common law jurisdictions, or legal systems that belong to the family of those jurisdictions, have more legal exclusion than we do have on the Continent – I mean, civil law systems. And the basic explanation of this difference is that they have jury trials; we basically have bench trials.
We have judges that rule on admissibility, evaluate the evidence and make decisions at the end, and also, after making a decision, they write down the reasons for which they made a certain decision.
As you probably know, as everyone knows, more properly, the jury doesn’t give you the reasons for which they decided in a certain way. The jury just gives you a verdict in a criminal case: guilty or not guilty. They don’t tell you why, right? That’s, I mean, how juries work. We can discuss about the merits and limits of having juries, but that’s another chapter of the discussion.
If you have a jury trial – and there are some reasons for having that, participation of the people in the administration of justice, blah, blah, blah – the problem is that there’s this sort of black box. You give evidence to the jury once it is admitted by the judge – remember, judge and jury. So, the parties ask the judge: can we use this evidence? Yes or no, depending on the rules of exclusion. And once you admit the evidence, you present the evidence to the jury, if it is a jury trial, and then the jury evaluates and the jury makes a decision, and they don’t tell you why they decided the case in a certain way.
So given that we don’t have a transparent outcome, because it’s a black box, you need more strict criteria at the beginning of the process, in the admissibility phase. At that phase of the proceedings, we need someone who scrutinizes the kind of evidence that is going to be presented to the fact finders, that is, the jury, and someone who is going to exclude the evidence if there’s something wrong with it, for some reason.
So, in continental systems, with professional judges – like in continental Europe, and I think in most countries of Latin America – the rules of exclusion are less strict or less stringent, because of the structure of the trial and the fact that it’s just the same decision maker from beginning to end, right.
For instance, the very, very large debate about scientific evidence that has dominated many discussions on evidence in the last decades is a debate that comes from the so-called Daubert trilogy – everyone knows that, I think, nowadays – which is a trilogy of decisions by the American Supreme Court about admissibility criteria of scientific evidence: what counts as scientific evidence or expert evidence, when it is admissible, when we can take an expert as a genuine expert versus, you know, junk science or just charlatans, when scientific research is serious, reliable, etc., and so deserves to be admitted and tried at trial.
And that discussion in the Anglo-American world has been about the admissibility criteria. Instead, in civil law countries, the discussion is mostly on the methods or criteria or rules or guidelines that the decision makers have to follow when they evaluate the evidence before making a decision. So, the evidence has already been admitted; in general, it is admitted, and then everything turns on how you evaluate it when you are a professional judge.
For instance, the Italian Cassation Court has made decisions that discuss that point. They take inspiration from Daubert and the Daubert trilogy in the US, but the procedural point is different. It’s not on admissibility; it’s on the evaluation of the evidence that was admitted and presented.
Bojan Spaić
Hmm.
Giovanni Tuzet
That’s an interesting thing that has to do with how trials are structured.
Julieta Rabanos
That’s a very interesting point, in fact. And I find it very interesting that you mentioned the jury as a black box, because that is something that we maybe were going to ask at the end about artificial intelligence and the possibility of using it in evidential reasoning.
But now that we have talked about admissibility criteria and we were talking about the black box, and then that maybe the inferences drawn from some specific, I don’t know, assessments or statements were not really transparent, maybe we could move to a question about evidence and inference.
Giovanni Tuzet
Yeah, yeah, yeah. This is something that I love because I have done a lot of work in this direction, I mean.
Julieta Rabanos
We came back to the abduction thing.
Giovanni Tuzet
Yeah, we go back to that now. We will discuss this.
Julieta Rabanos
Yes, finally.
Bojan Spaić
We were going all the way to Peirce.
Giovanni Tuzet
Evaluation and decision – but before evaluation and decision, you have inference, right. So, you are presented with evidence. Assume that you are a fact finder, judge or jury – from this point of view, it doesn’t make a difference. You are presented with evidence, and you have to infer something from it, right. Evidence doesn’t speak for itself, with some exceptions maybe. Yeah, testimony, in a way, speaks for itself.
But evidence, generally speaking, doesn’t speak for itself in the sense that when you are given a document or an object or an item that counts as evidence, you have to infer from it what the evidence is going, what the evidence is supposed to prove. So, you have to draw a conclusion from the evidence, or, to put it better, draw a set of premises that are the linguistic and conceptual translation of what you were shown.
You are shown an object – it’s not enough. You have to translate the object, the thing, the item that was shown to you; you have to translate it into an inference: some premises and conclusions that follow from it. That’s interesting, because it’s again a point that has to do with the inferentialism.
And the next question is: what kind of inferences do we use in law, in legal fact finding? Basically, my answer is: any kind of inference, any kind of inference is fine, to some extent. But mostly we reason by abduction. And abduction was the name that Charles Peirce, the philosopher I mentioned earlier and, in that post, gave to the inference that provides an explanation of some data, of certain pieces of evidence or certain information.
When you have to explain something, you perform what Peirce called an abduction. You do it in medicine, for instance, when someone has symptoms that have to be translated into, you know, medical concepts and categories, and you have to explain why the patient is like that, why the person feels pain, or whatever.
And it’s an inference that is hypothetical. When you perform an abduction, you make a hypothesis, you formulate a hypothesis on what’s going on – in medicine, in science, in law also, in history. You find some evidence, you have to reconstruct the case, you have to explain why those items were there, who left those items there, what happened, who killed the victim, who stole the goods or whatever. So, you have to perform hypotheses, and that’s the business of abduction.
When you have more than one hypothesis in play – so you have alternative hypotheses, you have to perform something more complex, which in the literature is called inference to the best explanation. You have some facts, some evidence that needs an explanation, and you have some rival hypotheses: A, B and C, or 1, 2 and 3, call them as you wish. You have to select the best among them. That’s called inference to the best explanation, which is a more complex inferential process than the simple fact of performing an abduction, which is the simple suggestion or the hypothesis explaining something.
When you have rival hypotheses, you have to select the best explanation according to some criteria that tell you what we mean by “best” in certain contexts.
In law, I mean anytime we have competing explanations of what we know, the fact finders need to select the best explanation. In the contemporary literature on law and legal proof in particular, I’ve worked on this and have a view of legal fact finding along those lines.
So, there’s abduction, there’s inference to the best explanation – which is a more complex inferential process – and there are also other forms of inference. For instance, you have induction whenever you have a generalization of data that is going to be used as a major premise in an inference explaining certain facts.
And this is a bit technical, I understand, but suppose that you need a premise concerning the specific fact and you want to draw a certain conclusion by providing an explanation of the facts that you are considering. Very, very often you need some general premise that allows you to move from one specific fact to another, that is, from the evidence to the other fact that explains the evidence, from the known facts to the unknown facts.
And the general premise of the inference is usually provided by some kind of background knowledge or some generalization that is part of our common sense, or our knowledge of the world, knowledge of how things usually happen in certain contexts, how usually people behave in certain contexts, etc., etc., etc.
So, there’s an inductive component of legal fact finding which is used anytime you have an inference with a major premise, and the major premise is the generalization that comes from our experience or some kind of science, when it’s more specific and it has to do with science, scientific stuff.
And, as William Twining said, generalizations are necessary when you have an inference that goes from one fact to another. You need a sort of bridge, a major premise that bridges the two. But at the same time, they are dangerous because generalizations are sometimes too, I mean, over-inclusive; they have biases in them, the prejudices that sometimes are embodied in generalizations, and so it’s a risky business. But, you know, we have to run that risk – I mean, in evidential reasoning we have to run that risk. We cannot avoid that. Deductive reasoning is fine, but there are no surprises in deduction. I mean, when you have a nice deduction, you have premises, and you draw something that is implicit in the premises and everybody is happy.
But with non-deductive reasoning, there are risks.
Julieta Rabanos
I am laughing because I am remembering when I was a student that there was a discussion about Sherlock Holmes – what type of reasoning was he doing, whether it was deduction, abduction, some induction going on.
Giovanni Tuzet
Unfortunately, people say that Sherlock Holmes performed deductions, and that’s wrong. That’s wrong. That’s wrong. In the novels, it is said that he performed deductions, but it’s wrong, technically speaking, logically speaking. They are not deductions; they are abductions and inferences to the best explanation, supported by inductions when you have a general premise that works in the inference.
Again, that general premise can come from science – can be scientific knowledge – can come from common sense, can come from experience. So, what happens is: in a certain context, usually, more often than not, and blah, blah, you see there are fuzzy boundaries there.
But there’s also room for deduction in evidence and fact finding. This may strike you because I just said that everything is risky and hypothetical, conjectural, so apparently there are no deductions. Well, maybe there are deductions in some cases. For instance, take the case of a conclusive alibi. If the crime was a crime that was performed in New York and you have conclusive evidence that the person was in Rome at the same time, same day, you can put that as a deductive inference; technically, it’s a modus tollens inference. It gives you the conclusion that necessarily that person did not commit the crime in New York, if it was conclusively shown that that person was in Rome, right. So, there’s also room for deductions in fact finding.
And to conclude, there’s also some room for analogical reasoning. This is something that I’m working on; I haven’t done much in the past on this. I have done work on analogy, but as lawyers traditionally do, I’ve worked on analogy in legal reasoning in the sense of reasoning with norms to fill gaps, for instance, in legal systems: draw analogies and blah, blah, blah.
There’s also analogical reasoning concerning the facts. For instance, there’s a causal process that is harder to reconstruct, and you try to reconstruct it using similar processes that you know better. There are toxic civil cases in which the plaintiff claims that the disease of the victim was caused by a certain substance. It’s unclear if that was the case, but we know that similar substances cause very similar diseases. You can use that as a sort of analogical inference that supports the hypothesis of the plaintiff.
Or in criminal cases, suppose that we don’t know the identity of the criminal, the individual who acted criminally, but we know exactly how the crime was performed – so we know the modus operandi, or the way the person committed the crime exactly – and we know from other cases that a certain individual committed in the past some crimes exactly in that manner. You can use that as an analogical inference; you can translate that into an analogical argument against the defendant. I’m not saying that it would be conclusive, I’m not saying that it will for sure show that the defendant is guilty also in this present case, but…
Bojan Spaić
We talked about some parts of this. So basically, when we talk about inference and evidence, to sum it up a bit: we basically start with an abduction, which we use to formulate the hypothesis, and then we go on to perform inference to the best explanation. And you have a couple of – as much as four – questions about this, to quote your son: basically, in any situation, as much as fifteen questions about your claims.
But I want to push a bit actually with another mode of inferring something about facts that has been the topic of much debate, especially in the literature on abduction. Now, I won’t push too much on this – whether abduction is better describing evidential reasoning, or it’s actually a prescriptive approach to legal reasoning which we follow in order to get logically valid conclusions from certain information or certain facts we call evidence, and so on.
But for quite a long time, the dominant approach, or one of the dominant approaches, was to represent evidential reasoning in terms of probability, or more precisely by using Bayes’ theory, to fit into the theorem the parts of evidential reasoning and to basically control the results of evidential reasoning, or even perform evidential reasoning. Could you comment on this relation between – and the debate regarding – probabilism or probability, evidential reasoning and abductivism?
Giovanni Tuzet
Abductive inferences provide hypotheses on the relevant facts. They are not necessarily providing probability values. But you can, to some extent, translate an abduction into an inference that gives you some kind of probability about what happened. Some abductions have stronger conclusions than others, and you can rephrase the point by saying that some abductions provide conclusions that have a probability which is greater than other abductions or other hypotheses.
So, there’s a matter of degree; it’s a matter of degree. And probabilists take this literally by saying: well, every time you have an inference which is non-deductive, because it has to do with uncertainty or limited information, etc., you have a conclusion that can and should be given a probability value – a probability value that is going to be calculated, for instance, according to Bayes’ rule. I’ll go back to this in a few minutes.
Abduction, on the other hand, can be taken qualitatively, as providing conclusions that are more or less plausible. In the recent debates, especially in the US – but now it’s an international debate – there are the so-called plausibilists, who claim that, as a matter of fact, fact finders do not come up with numbers at the end of their reasoning. I mean, they come up with conclusions that are more or less plausible or probable, in a sort of intuitive sense, using not only logic but also common sense, background knowledge of the world, and all the cognitive tools that we have and we normally use.
The probabilists want to be more specific and more precise, and they say: no, no, that’s too fuzzy, that’s too vague. You don’t understand exactly what plausibility means, what you mean by a plausible conclusion, when a conclusion is more plausible than another – that’s too undetermined. So probabilists want to be more precise and have probability values.
That’s nice if you love numbers in general, if you love precision, if you want to be rigorous in your thinking – that’s nice. The problem is that very often we don’t have sufficient data. That’s one problem. So, making claims in terms of exact probability values is very often quite artificial, because we don’t have data.
If we have statistical data, we can use Bayes’ theorem and other information quantitatively represented, if the data are reliable, and draw conclusions that are applications of Bayes’ theorem and give you the so-called posterior probability of a certain hypothesis. Let me say a bit more about this. Bayes’ theorem is a mathematical thing that allows you to move from prior probabilities to posterior ones – from prior probabilities to posterior probabilities.
You need the prior probability about the hypothesis being true. Then you consider the evidence and, more technically speaking, you consider the so-called likelihood of having the evidence if the hypothesis is true or having it if the hypothesis is false – this is, I mean, a kind of technical thing. And once you have likelihoods and you consider the prior probabilities and the likelihoods with the mathematical formula, then you have the posterior probabilities that represent the probability of the hypothesis, given the evidence.
When you have the prior probability, you don’t consider the evidence. When you consider the evidence – if it’s relevant, that is, if it increases or decreases the probability – you can, applying Bayes’ theorem, have the posterior probability, and that is fine. I mean, that’s mathematically fine, and nobody, as far as I know, is against the theorem itself.
The problem is that in legal contexts, legal fact finding, very often you don’t have precise data to start the process. And even when you have data that are OK, that are reliable, statistically fine, it can be very, very difficult to perform the calculation, especially if you have a jury, with laypeople who don’t know anything about Bayes’ theorem, probability theory, really. And so, they do not know.
So, they basically follow logic, common sense, everyday reasoning, etc., etc. – these things that lawyers have always exploited in the sense of arguing. Now, the probabilists say: oh, that’s the past, that’s too fuzzy, too vague, indeterminate. You don’t know what happens. If we have, on the contrary, a system that uses correctly probability theory and Bayes’ theorem, we will have more reliable results and more accurate results and more precise results. That’s the debate.
And to be honest, I rather side – I mean, I am a critic of the probabilists’ approach, because I think that, as far as I can understand, generally in litigated cases we don’t have the data that enable us to use sophisticated tools like Bayes’ theorem. And in cases in which we have data, statistical data, we have the problem that fact finders are usually unable to perform that thing.
So, the next step might be: well, let’s replace those human decision makers with machines, right?
Julieta Rabanos
Gasp. We arrived to the black box. But do you think that the replacement of these, or the possible replacement of these human fact finders, is more about, for example, coming with a hypothesis or the possibility of computing these probabilities, or about evaluation for decisions?
Giovanni Tuzet
You can have both. I think there are systems that generate hypotheses: you give data and ask the system for explanations, and the system provides you hypotheses. And there are systems that evaluate the evidence presented in favour of or against the hypotheses.
If a system uses Bayes’ theorem or so-called Bayesian nets, those systems make a calculation of the combined value of the evidence, because in complex cases you have a set of evidentiary items – you have testimonies, objects, documents, data of various sorts – and you collect, you have to collect, all those items and calculate the final posterior probability of the hypothesis, given all the stuff that you have as evidence.
That’s, mathematically in some cases, a very complicated thing. Humans are not very good at that, and some kind of machines – let’s use this term – would certainly do that better than we do.
The question is: do we want that? And, to go back to my previous point, are we sure that the data we use as input are fine? Are we sure that the system is bias-free? Are we sure that an artificial intelligence system is not designed in a way that incorporates biases of various sorts? You are familiar with that – there are discussions on the biases that are incorporated in some artificial intelligence systems. That’s a danger.
That’s not a conceptual argument against artificial intelligence systems; that’s a danger, because, of course, the probabilists or the supporters of artificial intelligence can reply: well, let’s build better systems, bias-free. Yeah, OK, in principle, that’s fine. But, you know, in human affairs – I mean, take a case, a kind of standard case – there are essential evidentiary items and issues that resist probabilistic or statistical translation, because they have to do with parts of our human life and things that are, I mean, that are not mathematical things; they don’t have mathematics in them.
Their nature and the ways in which human things happen are such that they resist, to some extent, formalization, mathematical calculation. So that’s a big problem, I think. Well, there are more details, but if you want to have a sort of overview of the thing, my impression is that we are going in that direction.
The use of artificial intelligence in law and in fact finding is growing in a sense; it’s expanding. You have it more and more and more, and so it will be in the future. We should be aware of the dangers of that and of the limits of that. I’m not against that a priori; maybe it can help us, can help us avoid some biases, overcome some limits of human reasoning.
But we should remember this: that we also have the problem of artificial precision. I mean, just giving numbers as such doesn’t tell you much. You want to have reliable numbers; you want to have calculations that in some way reflect the facts of the case – which is not that easy.
Bojan Spaić
And I somehow feel that you both jumped into the deep end of the pool with the idea. I think it was a sudden change of topic, so I’ll just try to come back for a second and try to make you discuss the various roles, because we were mostly talking about human reasoning or human judicial reasoning regarding evidence, then we passed to these formalized probability, these mathematical formalizations that might eventually help us or might in some way even be beyond our human abilities from time to time to compute, to calculate.
And so, we kind of jumped into AI, but actually there are various ways in which artificial intelligence figures in evidential reasoning. We haven’t talked much about various kinds of evidence, though I know that in philosophy of evidence law it is oftentimes an interesting topic, so we might touch upon it here.
Actually, as far as I know, AI also – and some theorists insist on that – also plays now an important role in gathering certain facts. For example, some of our systems, some of our administrative law systems – for example, speed on the highway, or registration of the speed on the highway, or various tests that we conduct, including blood tests, various DNA comparisons – are also in part done by artificial intelligence. Then we could say that artificial intelligence kind of figures as a source of evidence.
But what you were basically doing is talking about these possibilities of, or the limits of, AI reasoning and the strong points of AI reasoning regarding evidential reasoning. Could you comment on this idea of AI already being used as a source of evidence?
Giovanni Tuzet
Yeah, it’s already used as a source, you’re right, and that is a challenge to our legal categories, by the way – that’s an interesting point for lawyers – because traditionally we used to think of evidence as being either, let’s say, documents or objects, or testimonies by witnesses and those traditional things.
Now we have more and more what people call digital evidence, or information that is captured in the environment by artificial intelligence systems that have senses – not like the human senses, but they have a way of capturing information from an environment and translating it into something that is then given as a hypothesis or a probability value, etc.
That’s an interesting phenomenon. It’s fascinating, I mean, it’s fascinating. And for the lawyers it’s harder in some cases to understand how to categorize that stuff. There are cases in the US, for instance, in which information provided by artificial intelligence systems is treated as if it were a form of testimony, which is bizarre – as if the machine were a sort of witness, and the machine is telling you. And you know, there are systems, machines that have an artificial voice, so you can ask questions to the machine and the machine tells you, yes, it was like this and like that.
Is that a sort of testimony, the machine working as a witness? Maybe. Or we should just use the traditional categories of documentary evidence and say no, this is information which is very similar to what in the past we called documents. Or is it scientific evidence, so we need experts explaining to us how the machines work?
So, we should call the human expert and say: please explain to us how the machine produced this evidence, because we want to be sure that the machine is reliable, the outcome is bias-free, etc., etc. So, explain to us how that works. Or – this is another possible scenario in the future – the machine itself is able to explain to you how it works, so it’s a self-explaining machine.
Imagine that the jury doesn’t understand a thing and asks the machine: can you tell us a bit more about the way you reached that outcome? “Yeah, sure, now I’ll explain to you, blah, blah, blah.” Interesting, right? You have open questions.
Bojan Spaić
It also sounds like something that we have to do a separate podcast on. I have the recent fad in the last five days regarding AI because people are slapping AI on everything. So, the recent fad is actually something that’s been around for a year now, but it started doing this thing.
Google published in 2023 something that is called, like, a note-taking app, called NotebookLM. So what you basically do is upload certain documents, and now the option that has been included recently is that, when you have these documents, you can actually click, with a click of a button, use the artificial intelligence Gemini – the artificial intelligence of Google – to make a podcast in which two persons, artificial persons, are talking about the topics which you uploaded.
So basically, this format that we did is already replaced. They do a good job actually, probably better than I do – not better than Jules, but better than me – but rules next time we might just introduce an artificial host instead of me.
Giovanni Tuzet
Guys, I want to add a point that we forgot to address. I know that time is running out and we forgot about standards of proof, which is a key aspect of the story, a key part of the story.
Evidence is presented once it has been admitted, then it is inferentialised, as I love to put it, then it is evaluated, and in the end, you need to make a decision, right? But in order to make a decision, you need a standard of proof.
That is, you need to understand whether the evidence, as it was inferentially treated and evaluated, satisfies the threshold – let’s say, the threshold of proof, the standard of proof. The standard of proof can be understood as a sort of threshold.
Remember what we discussed about Jordi Ferrer’s view? We said there’s sufficient evidence. Sufficient means here the idea that you satisfy a certain threshold, a certain standard of proof that is needed to make a certain decision. And there are different standards in different contexts.
This is a point that traditionally was discussed in the Anglo-American tradition. On the Continent there’s less discussion, traditionally, on this, and nowadays people are starting – thanks, by the way, to the work of people like Jordi – there’s more discussion about this. Everyone knows about the “beyond a reasonable doubt” standard, which is the traditional criminal law standard in Anglo-American jurisdictions, right.
The idea is that the evidence must be such that it satisfies that very demanding standard, which is proof beyond a reasonable doubt – so the fact of being proven beyond a reasonable doubt; any doubt about that would be unreasonable. OK, that’s a very demanding standard.
In civil cases, traditionally, you have a lower standard, which the Americans call the preponderance of evidence, and in the English tradition, in the British tradition, is called the balance of probabilities. So it’s the idea, if you want to put it numerically, just for the sake of clarity – if you want to put it numerically, the civil standard, the preponderance of evidence, means that the plaintiff has satisfied the standard, so is entitled to a decision in their favour, if the story of the plaintiff is more probable than not, has a probability which is greater than 0.5.
Any probability greater than 0.5 satisfies the preponderance of evidence or balance of probabilities standard. In criminal cases, it’s higher, right. Beyond a reasonable doubt is a higher standard, more demanding standard. Numerically, we don’t know exactly what that means. Some people say 0.9 probability and beyond, or 0.95, or 0.97, or 0.99 – I don’t know – but it’s a more demanding standard.
What is the reason for having different standards? Here the answer is quite clear. The stakes are different. In the criminal case, usually the stakes are different from the stakes of a civil case, and we want to protect criminal defendants up to the point of using a more demanding standard like “beyond a reasonable doubt” because, I mean, being convicted for a criminal offence means much more than losing a civil case, right?
Because you will be in jail or even – I mean, you could even be executed if the system has capital punishment. So, we want a very high standard in order to protect criminal defendants. That is a problem, because that high standard is likely to generate more false negatives than false positives. So, you will have more guilty people free, but that’s the price that we want to pay to have less innocent people convicted, right.
And we don’t have that problem with civil cases usually. That’s the reason for having a lower standard, which is, traditionally, the view in Anglo-American systems. On the Continent things are not that clear. Why is that? Because traditionally we were more concerned with criteria or methods of evaluation of evidence by judges.
There’s a lot of discussion in continental systems about la sana critica and l’intime conviction in France and similar things on the Continent. So usually, we have less discussion on the standards of proof that dominated, on the contrary, the discussion in the Anglo-American world.
But nowadays things are not like that, because many, many continental jurisdictions have imported standards of proof. In Italy, for instance, in the Code of Criminal Procedure we have a positive indication that the standard of proof is “beyond a reasonable doubt”, and the same in other countries – one is Chile, one is Colombia, if I remember correctly. There are other countries that, as a matter of positive law or as a matter of judicial practice, have incorporated standards of proof in criminal cases.
In the civil cases, things remain less clear in continental jurisdictions. For instance, if you ask an Italian lawyer what the standard of proof in a civil case in Italy is, they will tell you: well, the satisfaction of the judge. I mean, you have proven a claim when you have persuaded the judge. Yes. That’s not very illuminating, you know.
Persuasion according to the preponderance of evidence, or persuasion according to anything else – persuasion by itself doesn’t tell much. OK, so we need to discuss it more.
Bojan Spaić
On the other hand, it might seem problematic that the standard in Serbia is certainty in criminal cases, which would lead you to believe – if you kind of look at the Serbian Code of Criminal Procedure – it would lead you to believe that there are, like, five to ten persons in jails in Serbia, because, you know, having this level of certainty. It says actually “certainty”, but of course the interpretation of the standard is…
They don’t really mean certainty; they mean kind of high degree of probability in any judge.
Giovanni Tuzet
Yeah, yeah, yeah. It’s not certainty, certainty literally speaking. In the past, they used the formula “moral certainty” – moral certainty, which is not mathematical certainty or scientific certainty, it’s moral certainty.
And in England and Wales, some years ago, there was an attempt – and now I’m not sure that things are going exactly in that direction – but there was an attempt to change the system and replace the traditional “beyond reasonable doubt” standard with a certainty standard: being sure. You know, lawyers routinely ask judges to explain to them what “beyond a reasonable doubt” means. What do you mean by that? When is a doubt reasonable? When is a doubt not reasonable? That’s the interpretive problem and conceptual problem of the formula.
And given that jurors very, very often are unsatisfied with that standard, because they have the impression of not getting the content of the words, well, let’s stop using that formula and say that you can convict only if you are sure – but not sure in that sense, “I’m really, really sure for all real selves”, not in the analytical sense, in the colloquial sense: “oh yeah, I’m sure”, not 100%.
Otherwise, you’re going to, of course, acquit anybody, almost.
Bojan Spaić
It has been a fantastic conversation. I enjoyed it a lot. I have two final questions. The first one: would an artificial intelligence convict OJ Simpson? The second one is: are you in Milano so that we can grab some drinks tonight in Navigli? Completely unrelated, completely unrelated.
Giovanni Tuzet
I’m not in Milan right now. I’ll arrive late tonight in Milan. If it’s not too late, I’ll text you and we’ll have a beer somewhere.
On the first question, OJ Simpson – well, you know the story. The problem was that the police tried to frame him, and that was one of the reasons for which he was acquitted. The other thing is that, well, there were also racial issues in the story, but probably an artificial intelligence system wouldn’t care about racial aspects but would have correctly excluded the evidence that was fabricated by the police.
So honestly, I don’t know. I don’t know what the system would do. But, although you know the story, let me mention another interesting fact: OJ Simpson was acquitted in the criminal case, but he was considered to be liable in the civil case about the same fact.
Because you know that the American system allows that: on the same facts, you can have a criminal case and a civil case, and in the civil case he lost, and he was considered liable. So, he had to pay damages to the relatives of his own wife, etc.
The legal explanation is this: well, that’s not surprising because the standard of proof is different, is lower in the civil case. So, the evidence that was not sufficient for a criminal decision of conviction was sufficient for a civil decision of liability.
A little footnote, by the way: it was not only that. In the civil case some additional evidence was presented, and in particular there was a picture that was found showing him with some shoes that he had asserted he didn’t have. Well, it’s a long story, but there was also some additional evidence.
Julieta Rabanos
Just a footnote on this footnote – not like the shoes one – but it is interesting what you mentioned about the civil and the criminal trials, because I know that in other jurisdictions it’s the same, but also if you are convicted in a criminal trial, immediately that translates the certainty, let’s say, of the proof that we are presented in the criminal trial to the civil trial.
So, you don’t have to do basically anything else. It is just: if you were convicted in a criminal trial, then naturally, because of this difference in the standard of proof, you are going to be liable.
Giovanni Tuzet
Yeah. And if there is what we call in Italy parte civile, that is, if in the criminal trial you have the victim defended by their lawyers and asking for damages, you can get compensation with that trial. You don’t need to go to a civil trial to get compensation for the harm you suffered.
So that’s an advantage for victims who don’t have financial means to, you know, have expensive lawyers, because the prosecution will do the job – the prosecution will do the job of, you know, showing that the defendant is guilty, and then you get compensated thanks to that. Of course, you need to have some lawyer, but the job is basically the prosecution’s job.
The problem is that the standard of proof is higher, so it’s harder to have a decision in your favour. In a civil case, you have a lower standard of proof, but you have to argue your case on your own. You don’t have the prosecution doing it for you.
Julieta Rabanos
Yeah. Well, I think that this is a rat.
Bojan Spaić
Absolutely. I just, I’m thinking now, looking at the time, we’re running at around one hour and a half, and since coincidentally I’m doing an entire week of lectures next week on evidence, I think that I’ll just play the students this podcast.
Giovanni Tuzet
All right, and next time we will also talk about poetry and… promised.
Julieta Rabanos
Yeah, that would be fantastic. And then we can ask the question whether we can replace musicians and poets with a…
Bojan Spaić
We have to.
Giovanni Tuzet
OK.
Bojan Spaić
But then we should mix it here. You don’t want to mix these two things, like just… just… just the dark side is calling to you, like he’s splitting all the time his two parts of his personality, the analytic philosopher and the poet. Come to the continental side.
Giovanni Tuzet
No, no. Two different podcasts: one about law and logic and the other about… poetry and music.
Bojan Spaić
Yeah, the host, a spin-off. The host should be kind of a sensitive guy like me, or Dracula.
Julieta Rabanos
That would be…
Bojan Spaić
Yeah, the whole…
Julieta Rabanos
But it’s a fantastic idea, like the side B of Philosopher Flow, Heavily Accented side B of Philosopher Flow. Great.
Bojan Spaić
Great.
Giovanni Tuzet
Actually…
Bojan Spaić
So if we ever make it, you have the first episode.
Giovanni Tuzet
Me too. Thanks, guys.
Julieta Rabanos
No, thank you for participating, and it was a fantastic conversation. So again, thank you very much for everything.
Here you can find the book that is mentioned in the podcast episode:
Giovanni Tuzet
La prova ragionata
Giuffrè – 2023
(*some minor grammatical and changes have been introduced in order to make the reading more fluid, but in no way altering the content or the format of each speaker’s interventions).
The HAPL podcast is powered by the EU Horizon Twinning project “Advancing Cooperation on The Foundations of Law – ALF” (project no. 101079177). This project is financed by the European Union.
