Through the conversation.
22 sectionsHosts Dave Pietraszewski and David Pinsof introduce guest Tadeg Quillien from the University of Edinburgh to discuss the complex and abstract topic of causality. The conversation explores the intellectual history of Edinburgh, referencing David Hume's empiricism and his foundational philosophical problems regarding induction, correlation, and how humans infer causal relationships from observations.
David Hume was concerned with why it is that we can come to have causal knowledge at all.
“David Hume was concerned with why it is that we can come to have causal knowledge at all.” 5:50
The problem of induction is really just that it's always correlation.
“the problem of induction, is it's always correlation” 10:06
The speakers discuss why distinguishing causation from mere correlation is a crucial fitness problem for organisms, rather than an abstract academic puzzle. Tadeg Quillien explains that relying solely on correlational predictive models is unstable when environmental interventions occur, such as installing a garden sprinkler that alters the statistical relationship between rain and wet grass.
Storing a predictive model of the world based purely on correlations is unstable.
“storing having a a predictive model of the world that is just a pattern of correlations um is not a very stable way of representing the world.” 11:17
Dave Pietraszewski, David Pinsof, and Tadeg Quillien discuss the limitations of correlational models in science and cognition, comparing them to fragile towers of blocks. They examine how the human mind constructs causal relationships rather than merely tracking sensory data, drawing parallels to how we project mental states or ownership onto the world.
as Paul Smaldino says, it's this brittle thing that as soon as you try to pick it up and move it, it's like a tower of blocks and it just falls apart.
“as Paul Smaldino says, it's this brittle thing that as soon as you try to pick it up and move it” 13:20
causality is not in the input itself, it's something that the mind constructs in the same way that you know the mental states of other people
“causality is not in the input itself, it's something that the mind constructs in the same way” 14:16
Tadeg Quillien explains how the human mind builds causal models composed of variables and asymmetric relationships, defined primarily through hypothetical interventions and manipulatory actions. The speakers discuss how people can easily reason about unexperienced scenarios, such as moving the moon or the sun disappearing, demonstrating that minds go far beyond simple associationism.
the relationship between two variables is asymmetric, right? So it's not just that we know that there is an association between clouds and rain.
“the relationship between two variables is asymmetric, right? So it's not just that we know” 17:03
0524:06 – 32:02Evolutionary Perspectives on Causal Cognition and Domain Generality
Tadeg Quillien, Dave Pietraszewski, and David Pinsof discuss whether the causal cognition literature is evolutionary, noting that it employs functionalist assumptions and Bayesian optimality analyses. The speakers examine the debate surrounding domain generality versus functional specificity in the mind, critiquing traditional views of domain concepts.
I think um I would say that the causal cognition literature is implicitly evolutionary
“I would say that the causal cognition literature is implicitly evolutionary” 24:37
We have systems that are functionally designed for the purpose of picking out causal relationships
“We have systems that are functionally designed for the purpose of picking out causal relationships” 29:02
0632:03 – 36:07Normative Reasoning and Causal Models in Diagnostic Tests
The speakers examine how people conform to normative standards of Bayesian reasoning, discussing classic psychological experiments by Tversky and Kahneman on diagnostic tests. Tadeg Quillien explains that participants perform much closer to Bayesian optimality when they are provided with a clear causal model explaining false positives.
It seems that people are able to engage in in good Bayesian reasoning, but they need to form a causal representation in order to do that
“people are able to engage in in good Bayesian reasoning, but they need to form a causal representation” 35:10
Guest Tadeg Quillien discusses the Markov condition in causal reasoning with hosts Dave Pietraszewski and David Pinsof, explaining how intermediate events screen off prior causes. They examine experimental findings where participants appear to violate normative rules by giving higher probabilities when additional non-causal factors are mentioned.
once you know whether it has rained or not, then anything that you learn about whether the clouds are there or not should not give you any more information
“once you know whether it has rained or not, then anything that you learn about whether the clouds are there or not should not give you any more information” 38:09
People will give a higher probability if you tell them that, in addition to the fact that there is rain, there are also clouds.
“People will give a higher probability if you tell them that, in addition to the fact that there is rain, there are also clouds.” 38:57
The hosts and Tadeg Quillien debate whether experimental violations of causal reasoning reflect true irrationality or rational adaptations to uncertainty, background assumptions, and computational constraints. Quillien suggests that apparent mistakes stem from approximation shortcuts or participants maintaining uncertainty about the experimenter's stated causal structure.
if you assume that people have this uncertainty and that they are averaging over that uncertainty, then you predict that the that they will make the kind of quote unquote mistakes
“if you assume that people have this uncertainty and that they are averaging over that uncertainty, then you predict that the that they will make the kind of quote unquote mistakes” 46:02
there's a different way that they could be wrong is that if the problem is super complicated then then of course you
“there's a different way that they could be wrong is that if the problem is super complicated then then of course you” 47:43
0948:09 – 52:25Probabilistic Inference and Social Welfare Trade-Offs
Host Dave Pietraszewski and guest Tadeg Quillien discuss experimental research on probabilistic inference in social domains, specifically how people infer others' welfare trade-off ratios and how these inferences predict emotions like gratitude and anger. Drawing on Aaron Sell's work, the speakers explain that anger is driven by perceived disregard rather than absolute payoff losses.
People are good at following normative principles of probabilistic inference in the social domain when inferring others' welfare trade-off ratios.
“that people are relatively quite they're quite good at following the normative principles of probabilistic inference” 48:12
The best predictor of gratitude and anger is the inferred welfare trade-off ratio rather than the payoff amounts or opportunity costs.
“the best predictor of how angry or grateful the people feel is not the best predictor is not how much they get” 50:24
Host David Pinsof and guest Tadeg Quillien examine the different meanings of the term "computational" across psychology and other sciences, distinguishing between mathematical/statistical tooling versus theories of information processing in the brain. They also contrast formal mathematical modeling with genuine mechanistic explanations of cognitive processes.
There are two distinct meanings of computational in psychology: using math or computers as research tools versus studying information processing performed by an organism.
“it's a bit of an unfortunate fact about the history of psychology that you have just two different meanings of computational.” 53:16
A formal mathematical model can sometimes describe observed variables without actually explaining the underlying information processing in the mind.
“then you're not actually really saying anything about information processing in the mind.” 59:06
111:00:10 – 1:08:40Quantitative Models, Simplicity, and Bayesian Inference in Mental Systems
Dave Pietraszewski, David Pinsof, and guest Tadeg Quillien discuss the limits of complex quantitative and structural equation models in science. They compare these to simple computational models, such as the evolutionary design of anger and gratitude, and explain Bayesian inference as a way to quantify uncertainty and beliefs using probabilities.
when you think about the design of anger, it's shockingly simple. You know, you're trying to figure out what cost a person would be willing to inflict on you if i if in exchange for whatever benefit to them.
“when you think about the design of anger, it's shockingly simple” 1:01:57
the best example of this is the theory of evolution itself. You know, it's extremely simple. You can lay it out in a couple sentences.
“the best example of this is the theory of evolution itself” 1:04:21
The speakers examine how psychological design narrows down complex causal chains to actionable points of intervention, discussing examples like forest fires and human decision-making from an evolutionary perspective.
a notion of causation is also nar it's not just all actual causation, it narrows down to like as David describes, sort of like points of intervention
“a notion of causation is also nar” 1:09:52
this is one of those problems that we solve so effortlessly and unconsciously that it just does not register that that we are doing it.
“this is one of those problems that we solve so effortlessly and unconsciously” 1:11:13
Dave Pietraszewski, David Pinsof, and guest Tadeg Quillien explore the problem of causal selection, discussing how humans intuitively select specific causes—like a lightning bolt rather than oxygen—for events. They examine how this cognitive process links to moral psychology, moral responsibility, and how people evaluate causal connections in intentional harmful actions.
So, this is called the problem of causal selection.
“this is called the problem of causal selection” 1:12:51
And so a lot of what is going on is this that some counterfactual possibilities are just more relevant than others.
“some counterfactual possibilities are just more relevant than others” 1:17:06
The speakers discuss what makes certain counterfactual possibilities relevant during causal attribution, considering probabilities, social norms, and interventionist accounts. They debate whether social expectations directly influence our judgments of responsibility or if they simply reflect efficient targets for future causal interventions and communication.
And then I mean where it gets maybe a little bit more spooky is that kind of social and and moral norms also seem to have an influence on what we think what or the the relevant kind of kind of actual possibilities.
“social and and moral norms also seem to have an influence” 1:18:14
I don't like it relevance language because I find that obscuring. Because I think it's sort of an outcome of the interventionist calculation.
“I don't like it relevance language because I find that obscuring” 1:23:59
Guest Tadeg Quillien discusses with hosts Dave Pietraszewski and David Pinsof how people attribute causes to events, using examples involving exams like history and math. Quillien explains that the most relevant counterfactual is not always chosen as the cause, and describes experiments showing how varying task requirements shifts causal intuitions.
the relevance of something is not there is not always a direct a direct mapping uh between the relevance of a counterfactual and whether it is selected as the cause
“the relevance of something is not there is not always a direct a direct mapping” 1:24:28
when you ask participants uh why did Alice graduate, they will tell you, Well, it is because she passed the history exam
“why did Alice graduate, they will tell you, Well, it is because she passed the history exam” 1:26:47
Tadeg Quillien outlines a parsimonious explanation where people simulate multiple counterfactual possibilities in proportion to their relevance and compute correlations to make causal judgments. The hosts and guest also consider whether these simulations occur unconsciously or via evolved cognitive shortcuts.
when people make a causal judgment they they imagine not just one um counterfactual possibility but they they simulate many of them
“when people make a causal judgment they they imagine not just one um counterfactual possibility” 1:30:23
there might be a way that our brain has also evolved some shortcut way of computing what would be the outcome of such counterfactual simulations
“there might be a way that our brain has also evolved some shortcut way” 1:35:05
Dave Pietraszewski, David Pinsof, and guest Tadeg Quillien discuss how causal reasoning operates through implicit models and learned cues, comparing automatic physical predictions with conscious simulations during chess games, and noting how explicit communication can influence counterfactual simulations.
it's an example, it's an illustration of the fact that the model itself is really couched at the kind of the David Mars computational level of of analysis
“it's an example, it's uh an illustration of the fact that the model itself uh is really couched at the kind of the David Mars computational level” 1:37:23
if you are in a novel situation, maybe this is the first time that you've been playing this chess... then you need to engage in explicit simulation
“that if you are in a novel situation, maybe this is the first time that you've been playing this chess, this particular kind of chess game.” 1:40:35
The hosts and Tadeg Quillien explore theory of mind, examining the evolutionary importance of social cognition, the challenge of attributing mental states like desires and beliefs, and how meta-representations and false-belief tasks function in predicting human behavior.
when I attribute to you mental states, then the richness of my explanation of your behavior is going to be much more effective
“whereas when I attribute to you mental states, then um the richness of my explanation of your behavior is going to be much uh much uh much more effective.” 1:43:54
a standard test for whether someone has meta representations is whether they are able to imagine that someone might represent the world in a different way than your own
“a standard test for whether someone has meta representations is whether they are able to imagine that someone might uh represent the world in a different way than than than your own.” 1:45:15
Host Dave Pietraszewski and guest Tadeg Quillien discuss the cognitive architecture of theory of mind, arguing against the naive view that humans continuously run full simulations of other brains. Instead, Quillien explains how agents use a simpler model featuring a physical representation of the world combined with binary tags for shared knowledge, balancing predictive power with memory constraints.
the idea that we have been putting forward is that you don't always need meta representation. And in fact in in most cases you you don't actually need it.
“you don't always need meta representation. And in fact in in most cases you you don't actually need it.” 1:49:35
when we did that, we find that indeed if your agent is sufficiently constrained in terms of its information processing capacities, then it will use a strategy that is very similar to the one that I just described
“if your agent is sufficiently constrained in terms of its information processing capacities, then it will use a strategy that is very similar to the one that I just described” 1:54:36
David Pinsof and Tadeg Quillien discuss whether minor tweaks or presets to an existing self-simulation engine can explain how we predict others' behavior. Quillien explains that sharing the same physical world makes this strategy efficient, and outlines how optimizing these computational models can generate predictions about non-human primates and tasks like the accidental true belief task.
by virtue of the fact that you're living in the same world, you will end up creating representations that are actually quite similar.
“by virtue of the fact that you're living in the same world, you will end up creating representations that are actually quite similar.” 1:57:48
one prediction that follows from that is that they will behave in a certain way in tasks that are a little bit like a false belief task, except that you have changed one thing.
“they will behave in a certain way in tasks that are a little bit like a false belief task, except that you have changed one thing.” 2:00:03
The hosts and Tadeg Quillien discuss how simulated agents and non-human primates handle complex scenarios with limited information processing capacity. Following this, Tadeg shares his intellectual journey from political science and David Hume to cognitive science and evolutionary perspectives under Pascal Boyer.
it turns out that this is exactly how monkeys behave if you give them uh that that kind of task
“it turns out that this is exactly how monkeys behave if you give them” 2:01:44
I was kind of blown away by the fact that oh this guy was asking just deep questions
“I was kind of blown away by the fact that oh this guy was asking just deep questions” 2:03:25
As the episode concludes, the hosts thank Tadeg Quillien for appearing on the show and playfully encourage listeners to leave reviews and support the podcast on Patreon by framing them as causal interventions on the world.
we're just giving you ways you can intervene on the world.
“we're just giving you ways you can intervene on the world.” 2:06:19