The DCIDE Framework published in Biological Reviews

New paper🎉 just published🎉 in Biological Reviews 🎉with Adrian Jaeggi. It has two major aims: 1. Introduce a new gold standard for evolutionary inference avoiding ‘just so storytelling’; 2. Apply it to autism to provide the strongest analysis of feasible evolutionary explanations for autism to date.It’s been 9 years (!!) in the making. Summary 🧵

First, here’s the link: https://onlinelibrary.wiley.com/doi/10.1111/brv.70010

Since the first night I stumbled upon evolutionary psychiatry, with a background in philosophy of science, I saw a major problem: how can you prove or decide between any of the (many) evolutionary hypotheses of a particular disorder?

This is the ‘just so storytelling’ criticism of evolutionary sciences generally. but especially troubles evolutionary psychology & psychiatry. It goes back to Darwin! Evolution is a historical process – what can we repeatedly observe today to know what happened in the past?

Since the mid 20th century, evolutionists have used ‘reverse-engineering’. Basically, by observing a functioning mechanism, you can infer what its function was in the past, and thus why it evolved to have this form.

This works for simple adaptations like eyes – eyes are so good at seeing, we can confidently infer they evolved to see. Same goes for lots of obvious physical and psychological adaptations. There’s often simple experiments and no debate.

Things get trickier when the traits get more complicated. The function of hunger is less controversial than the function of jealousy. Then low mood is trickier…. And then what about depression?

The problem is, we can run modern experiments to try and test the function of jealousy, low mood or depression, but the interpretation of those experiments could often be taken many ways – they don’t narrow down hypotheses sufficiently. And contradictory evidence can often be raised…

We get in a mess of citing e.g. how depression helps disengage from conflict (a function!) but also causes cognitive slowing (a dysfunction!). Researchers can cherry pick findings which fit a functional story and tell a ‘just-so’ story about how it all fits together. Others can pick different evidence and fit it together in a different story.

This has been the state of the art since Darwin. It hasn’t really troubled more hardcore evolutionary biologists who can e.g. study reproductive success of study systems over multiple generations, and engage in invasive experimental designs. It has definitely troubled evolutionary psychology. Evolutionary psychiatry is doubly screwed.

Why is evolutionary psychiatry worse off? Firstly, because the targets of explanation are mental disorders, and reverse engineering seeks observation of pristinely ‘engineered’ function. Mental disorders are probably not pristinely functioning. Certainly not sufficiently uncontroversially so to allow agreement like ‘hunger is for seeking food’.

The problem with mental disorders is they could be related to adaptations, and need some kind of evolutionary explanation, but in a complex way: they could be adaptations going haywire in some individuals, or the trade-offs of some other functional trait, or manifesting weirdly due to modern environments.

(actually I have another paper, currently a pre-print, on the extent of these possibilities and precisely defining ‘dysfunction’ and different types of evolved disorder: https://doi.org/10.31234/osf.io/qbgke. For another time…)

The other huge problem is that initial mental disorder labels are poorly defined. ‘Autism spectrum disorder’ can include some of the richest people in the world, or some of the most disabled. There are multiple different causes of autism, or depression, or schizophrenia etc etc.

This is the challenge. But it’s one that needs to be faced. Psychiatry is in a mess. It can’t handle this complexity, especially with widening diagnostic labels which are incorporating more and more people nearer to ‘normality’.

An evolutionary approach is the only one which can pick apart the functional from dysfunctional and start making sense of these complexities. But doing so needed a more rigorous, systematic approach to evolutionary analysis than previously existed

This is the point of the DCIDE framework. It lays out five steps for evidence analysis and interpretation which can address this complexity and provide, for the first time, a set of standard principles for comparing and evaluating the sufficiency of evolutionary hypotheses.

There are five stages: Describe, Categorise, Integrate, Depict and Evaluate. The first three (DCI) deal with problems common to evolutionary psychiatry and medicine in particular: identifying the right target of explanation (maybe we shouldn’t be explaining ‘schizophrenia’, but the psychosis spectrum more broadly; not depression, but low mood). The second two (DE) concentrate on rigorous evaluation of competing hypotheses.

Like in the paper, here I’ll run through each stage and what it looks like exemplified with autism

Description is the first necessary step – identify what you want to explain. Autism diagnostic criteria have changed over time, making this messy, but right now have two main criteria, of i) communication and social interaction differences and ii) restricted or repetitive behaviours and interests.

A big problem is how broadly this can be interpreted. Restricted behaviours can be obsessions with maths or physical self-harm and head banging. Social differences; discomfort with eye contact or complete inability to speak. Inevitably, many different things can happen to cause an ‘autism’ diagnosis

The DCIDE framework can handle this, though! Evolutionary theory is the perfect (and really, only) way to partition these different cases into strictly categorically distinct causes.

This starts in Categorisation, which identifies cases which definitely don’t need an explanation referencing evolutionary history, removing them from the analysis, explained as simple dysfunction

This is done with biomedical variables, asking the question: was this inherited? If not, we don’t need an evolutionary explanation. Neuroscientific, genetic and environmental evidence can find many cases like this.

For autism, this means identifying up to 20% of cases (although probably less with recently expanded diagnostic criteria), mostly caused by de novo mutations, mostly with more severe intellectual disability and co-occurring conditions. These don’t need adaptive explanation.

However, the majority of cases can’t be explained by these simple, classically ‘pathological’ processes. These are the ones which remain a mystery to psychiatry and medicine. And they need evolutionary analysis.

So far, pretty simple, and nothing particularly novel. Integration is where it gets interesting. This is also where one of the major problems with naïve evolutionary psychiatry arises and needs addressing…

Coming into the field as a novice, you might think the aim is to explain a disease with an evolutionary (probably functional) explanation. Schizophrenia – whats its function? Eating disorders – what’s their function? This is the most common mistake we see (and I’ve made it myself).

This attracts (valid) scepticism and pushes people back towards simple disease ideas. But this is missing a major middle solution (which I think can help solve basically all psychiatry)! There are many ways in which adaptive functional processes can cause traits which are not in themselves adaptations – as by-products (image from preprint, not DCIDE paper)

The obvious example of this is sickle cell disease as a by-product of protection from malaria. The homozygous population suffer, but heterozygotes benefit. Essentially all common mental disorders are related to ‘subclinical’ symptoms in a wider population. And we often diagnose by the harmful symptoms (autism disabilities) rather than associated benefits (autistic strengths) which leads to a skewed lens on the people who are worse off.

Here’s the critical thing: by-products of adaptations shouldn’t look like simple pathologies, confusing biomedical research, despite often being quite harmful! They still need an evolutionary explanation, but we might be focussed on the wrong people and traits! Akin to trying to understand sickle cell without looking at the heterozygotes.

(My bet is that this is the fundamental reason we have spent billions and decades with basically zero progress in psychiatric research – we’ve been applying simple pathology models to try and study by-products)

So how do we Integrate the appropriate target for evolutionary explanation? Looking for positive characteristics (e.g. strengths of autism) might be one way, but that could lead to storytelling again, and is unprincipled and not theory driven. There's a better way.

Visibility to selection! Evolution acts upon phenotypes based on how ‘visible’ they are throughout life, and throughout the (ancestral) population. Assessing the broader effects of the inherited tendencies outside of the diagnostic label is necessary, and principled.

Using these ‘visibility variables’ we can derive standard, universally applicable criteria to try and extend our analyses to any related trait which has a better chance of being adaptive because it’s more visible. Like extending to heterozygotes of sickle cell.

When you analyse autism, you find it is highly visible in age of effect (early), duration (lifelong) and environmental plasticity (it basically arises regardless of environment). However, it is defined (or Described) in such a way to ignore female and broader spectrum (the ‘Broad Autism Phenotype’) manifestations.

So, if we want to ask questions about adaptive processes leading to autism, we should Integrate female and subclinical cases. It’s possible the male cases and extremes are costly by-products of subtler adaptive manifestations we more rarely diagnose...

My interest in evolutionary psychiatry has always come downstream of recognising a longstanding paradox: why heritable mental illness persists in our species. After Integration the ‘paradox of common, harmful, heritable mental disorders’ is refined into the paradox of mental disorders with early onset, long-lasting, ancestrally activated, common, heritable correlates, with no discoverable pathology!

Now we need to start thinking about adaptive explanations: this starts in Depiction. Basically, present hypotheses to explain why these traits persist through the generations – why they are reproductively successful. Often this will concentrate on evidence of the ‘function in action’: the wing flying, the eye seeing, similarly to reverse-engineering.

Importantly, any account, including non-evolutionary ones, can be presented here: the question they should all try and answer is why the inherited tendencies persist. Microbiome, trauma, chemical imbalance, evil spirit possession: whatever, just present a hypothesis to explain the Integrated trait's persistence down the generations.

Leading evolutionary hypotheses for autism come from Baron Cohen, Del Giudice and Crespi. They have similarities and differences.

Basically, everyone agrees that the strengths observed in autistic people and their family members (in specific abilities, reasoning, attention to detail, memory and more) must have something to do with why autism persists. They differ on details of the model.

In Evaluation, we start to test these. The first two steps are pretty standard scientific practise: i) asking whether evidence, theory and methods are sound. ii) checking that one account doesn’t refute another straight out. The third step (‘sufficiency’) is the more interesting contribution of the DCIDE framework.

In Evaluating the ‘reliability’ of the three accounts, you find that Del Giudice and Baron Cohen’s hypotheses are very similar – and really, the unique parts of Del Giudice’s hypothesis about different relationship styles is quite poorly evidenced. For simplicity, they can be merged. This leaves the ‘systemising social niche specialisation’ hypothesis facing the ‘high intelligence by-product’ hypothesis of Crespi

Basically, the systemising niche hypothesis states that autistic traits are ‘specialised minds’ (see mine and @ajaeggi other paper). Cognitive strengths and weaknesses balance out, with autistic traits advantageous in a proportion of the population because of their benefits to systemising, but don’t spread universally because they have costs.

The by-product hypothesis, on the other hand, points to intelligence in the whole human species as the advantage: autistic traits are harmful overshoots which are exaggerated versions of ‘perceptual’ intelligence. This doesn’t imply autistic traits were somehow selected for strength/weakness profiles.

So, how to compare these hypotheses (and any others) in a systematic and structured way, avoiding cherry-picking and making clear which ‘story’ is actually the best explanation?  Well, we can derive standardised, theory derived implications from a set of circumstantial facts observable for any given trait… such as visibility variables!

For example, does a trait arise in infancy or adolescence? Does it arise in everyone, or only 10% of people? Does it arise in certain environmental circumstances? These are facts that can be asked of every trait, and which the correct explanation must answer – whether microbiome, mutation, evil spirit possession, etc.

Functional evolutionary perspectives can perfectly fit visibility variables – or not. Whatever languages’ explanation, it should explain all these facts. Whatever depressions’ explanation, it should explain all these facts. Here Depictions are tested: do they really perfectly explain the evidence, or do they have to be shoehorned in like ‘just so’ stories?

Similarly, any by-product or pathology (or whatever) hypothesis should explain all these facts. They are simple, observable, easy to agree upon objectively, with standardisable inferential implications.

You can’t cherry pick these variables. You could publish a fifty thousand word review with thousands of citations on the costs of depression; and another of the benefits. You could run a thousand questionnaires asking about autistic strengths and weaknesses. They don’t have the same standardisable, theory-derived interpretations.

So what happens when you Evaluate the competing Depictions ability to explain these variables? Well, basically, Crespi’s by-product model struggles, whilst the systemising social niche specialisation hypothesis aligns with most of the facts

The specific visibility of autism (early onset, lifelong, distinct male and female manifestations, subtly environmentally variant, common enough in the population to be in every ancestral tribe) isn’t what we predict from by-products, but is exactly what we expect from functional adaptations of the type suggested by Baron Cohen and Del Giudice

Note that if any of this evidence went the wrong way: if autistic traits were rarer, or late onset, or didn’t adapt to environments in the same way, the social niche specialisation hypothesis could be excluded. But the evidence pretty sufficiently aligns.

Also note that whatever the correct explanation, it has to explain this evidence. Microbiome, by-product, mutations, spirits, whatever… during Evaluation, every hypothesis faces these same areas of evidence, with standardised interpretations derived from basic evolutionary theory. Why did evolution leave us with traits with these characteristics?

We conclude that it would be surprising if a very different hypothesis has led to this exact evidence. Could it all be very conveniently arising disease or mutations, lending strengths, arising at the right time, in the proportion of the population that means every ancestor knew and interacted with autistic people? Sure… maybe. But that’s quite a stretch.

Until direct evidence of such pathogenic processes is found, the systemising social niche hypothesis of autism best fits the totality of the evidence. If (and when) new evidence arises, or old evidence is dismissed, the DCIDE framework will allow us to reinterpret these conclusions.

At the end of the paper, we note that this could help evolutionary psychology move past ‘just-so’ criticisms. We show how it justifies a leading evolutionary explanation of jealousy. Much of the evidence is already out there, the theory is already agreed. If looking to gather more data, the DCIDE framework points to what is missing. When incorporating new findings, it provides structure. Evolutionary psychology and psychiatry could start making substantial progress.

So much more is available to support scientific inference of evolutionary processes than in Darwin’s time. We have better theory, evidence and computational power than in the 20th century. Evolutionary hypotheses can be triangulated upon from multiple inferential angles. We can start agreeing upon these long-sought solutions to huge scientific and social problems.

We finish the paper by noting that psychiatry is plagued by a distinct problem of pathogenic storytelling – for decades (centuries? millennia?) we have concocted stories of what might have gone wrong – in biology, in the family, in the spirit. These speculations are often based on very little – an experiment or two, and a lot of conjecture about pathology. We think they might have been ‘just-no’ storytelling. /Fin