Netflix, True Crime and the Illusion of Reason

The Dreamer

You may say I’m a dreamer, but I’m not the only one…
- John Lennon

Kant would not have liked John Lennon. He had a bit of a problem with dreamers and not the ambitious five year plan type of dreamer, he’s after the actual dreamers.

In his account of reason a distinction is made that our senses do not have the capacity to make mistakes. Sensations do not make claims about themselves, it is only when judgement is made about the experience that mistakes can start to arrive. Let’s take an easy example of a dream in which you win the lottery. The experience, the dream, happened and there is no error in having the dream. Mistakes then happen when waking from the dream and instantly resigning from your job, book a luxury cruise and explain to the bank manager not to worry and that funds will be sent from Camelot soon. The judgement about what the experience represented was wrong.

Kant reaches the problem of how we distinguish an actual experience from imagination, dream or delusion in one of his highly accessible, simple to read and understand, seminal works the Critique of Pure Reason. It’s not an answer that something simply feels more real as that falls under the same trappings of judgement, no, a putative experience has to be tested through its particulars and through its coherence with the conditions of actual experience. Would you like a drink to wash down that word salad I just served? Here’s a tall glass of claims about experience have to hang together with the wider structure of actual experience.

Let’s build on the lottery dreamer from earlier. Suppose, for example, they believe they receive psychic visions and start to receive visions about an ongoing murder trial. The vision itself is just an experience they had, maybe a dream or an intrusive image, something they sincerely cannot explain. Enter judgement to bang the gavel and declare that the dream revealed something that actually happened, well now it has to connect with the rest of experience. Does the dream fit the physical evidence? Does it fit the timeline? Does it fit observations made independently by other people? Does anything outside the dream provide a reason to believe it corresponds to an event in the world? Seems straightforward that when these connections all crumble away one is left only with the dream. Writing in the 18th Century, though, Kant couldn’t possibly have conceived of anything like TikTok.

With TikTok and other social media also comes an interesting problem because our dreamer is able to take their vision to a connected but distant society where they inevitably find others who believe the same thing, or people willing to believe anything when said with a sneering tone and over exaggerated hand movements. One turns into two, three, fifty, two-thousand. Videos are made for ten thousand more people to watch and flood comments of reinforcement that they had been thinking exactly the same thing, it spreads between creators and then lands back in the algorithm of the original dreamer. Their belief is suddenly wonderfully coherent with the judgements of thousands of other people.

“Look at all this corroboration” they will cry…

Except it isn’t corroboration.

Ten people independently reaching the same conclusion from different evidence might suggest something. Ten thousand people repeating a conclusion they were served in the same algorithmic ecosystem only tells us that the conclusion travelled well. Question why it travels well and the answer is most likely due to the salacious nature of what is being discussed. It is a meaningful difference though and one that is being ignored in favour of a desire to feel like they have turned the case wide open. Kant’s coherence is not an applause meter for some community centre’s talent content. Agreement between people only becomes useful where those judgements are themselves constrained by a shared world. If all of the apparent confirmation is ultimately descended from the same unsupported assumption nothing has really been confirmed; it only looks like it has. We have created the illusion of reason.

…“Oh, just one more thing”

I’ll make something clear before going further. I like true crime, quite a lot actually. My primary interest has never really been the spectacle or gruesome nature of the crimes (though the more depraved they are the more interesting they are). What fascinates me is that every person involved remains human. And as someone who claims to be human, being made up of pretty much all the same parts makes one question their own capacity for such things and what it is that changes. The mind is capable of extraordinary variation across the whole spectrum of things like intelligence, compulsion, personality, childhood, inhibition, trauma, social conditions, neurological different, opportunity; all of it colliding differently in different people. Someone like Ed Kemper can sit in interviews years after his crimes and appear disturbingly capable of analysing parts of his own behaviour. How much of that awareness existed at the time and how much was constructed retrospectively? If someone recognises an urge as destructive how much control do they actually have over it? Where is the line between explanation and responsibility? All very interesting questions to me and evidently to quite a few other people too. A 2024 Edison Research study estimated that 84% of Americans aged 13 and over consume true crime material through TV, podcasts, YouTube and other social media or a combination of them. Pew also found true crime to be the single most common subject among the top ranked US Podcasts it studied. Perhaps most interesting was the statistic on the reasons people gave for listening, where half of true crime podcast listeners said they liked participating in solving the mystery.

We aren’t always content to watch, we want to have a go ourselves.

True crime and TikTok did not create this phenomenon. In the nineteenth century people obsessed over Lizzie Borden. It was the media that actively shaped the mythos around Jack the Ripper and turned the case into a product of an expanding mass press. The serialisation of murder for public entertainment and profit has a long and sordid history a hundred years before Netflix even hit the scene. What has changed is the machinery around it. Now mystery is mainlined direct into the vein of the true crime junkie, observation turns to involvement as they search to find the buried plot twist; fiending for the next crumb of perceived evidence, forming their own Reddit based opium dens and happily loading and lighting the pipe for the next person who wanders into their “truth”.

This new machinery and mechanism made it easy to ignore evidence and replace it with intuition and vibes and not have that opinion ridiculed. I would argue that the greatest detective of all time is Columbo, just look at the numbers. He solved at minimum 69 crimes, they just televised his best 69 so think about the untold number that he solved in between them. I joke, of course, the point is that Columbo was not a classic “whodunit” format of story. We saw who did it at the start which then made the entertainment and mystery around “how will Columbo work it out?” as he inevitably will from the smallest most innocuous detail. The crime is not presented as a puzzle there to be solved and even in the more traditional presentations such as Holmes or Poirot there are is at least a consistency in the reaching of a conclusion. Importantly, by the end of the story there were no questions left unanswered and the mystery had been solved.

And then December 2015 happened.

Making a Detective

Hands up, I will admit I can be a contrarian at times and mostly around popular media. I didn’t engage with Making a Murderer at first but with what is an almost offensively compelling premise and narrative it only took a few months for me to come around. On 18 December 2015, Netflix released all ten episodes and the landscape of true crime shifted seismically.

Steven Avery spent 18 years in prison for a sexual assault he did not commit. DNA evidence did eventually exonerate him and after his release he pursued a multi-million dollar civil action lawsuit connected to the failures surrounding that conviction. Then, two years after his exoneration, Avery was arrested and subsequently convicted alongside his nephew, Brendan Dassey, for the murder of photographer Teresa Halbach. An already extraordinary sequence of events and that’s before any cameras were introduced, which Netflix were happy to provide.

The series arrived with every ingredient required to make suspicion of the events and outcome feel reasonable. There had been a genuine miscarriage of justice resulting in a man being wrongfully imprisoned for a significant portion of his life. In that case there really were legitimate reasons to scrutinise how local authorities had previously treated him. Then, while the litigation that arose from that history was still weighing over events, the same man happens to find himself accused of an appalling new crime. The CIA couldn’t have designed a more fertile ground for conspiracy theory claims to be cultivated. Importantly, Making a Murderer did not present a closed historical event where every uncertainty had subsequently been resolved. Avery and Dassey were still in prison. Questions remained contested. Unlike Columbo where we are invited to uncover the investigative process, or even Poirot where we are invited to deduce along with the detective; with Making a Murderer we are repeatedly but not explicitly invited to directly investigate. There was evidence to examine, investigators being questioned, alternative explanations to consider and ultimately the viewer was left to then decide for themselves which version of events made the most sense. Dreamed or otherwise.

It wasn’t just another true crime documentary, it was a cultural event. Netflix did not publish official audience numbers but there are estimates that have put the US audience at more than 19 million within the first 35 days. The analytics also suggested at the time that the audience grew through word of mouth and as people started to talk the viewer base reportedly doubled between its first and third weeks. People weren’t being recommended the documentary series to watch, they were being recruited into the investigative team from a recruitment process that started by being given a run down of all the unfounded claims and questions left unanswered. And somewhere in the approximately ten hours between pressing play and hearing Avery’s voice crackling through another prison telephone line, an extraordinary number of thirty year homicide detectives apparently qualified overnight.

I might have been one of them.

I was enticed by the trick that they play.

Notice that the first series of ten episodes was approximately ten hours of viewing. Sounds like a long time really but in the context of this narrative, is it? In the narrative of any true crime case, is it enough? A documentary necessarily compresses reality. Hundreds, maybe even thousands of hours of events, documents, interview and proceedings become a narrative that a human can actually sit and follow. I’m not calling it dishonest, merely highlighting that what the view receives is not the evidence. It is a presentation constructed from evidence and subject to the same bias as anyone else in that presentation. There are facts of what definitely happened which are then tied very closed to the question of “what do I think about what I just watched?” which risks conflation of evidence with feeling.

Making a Murderer didn’t invent armchair detectives but it did arrive at exactly the right cultural moment to demonstrate how easily true crime could become participatory. Ten years later, we seem to have now perfected the format.

The Trial That Isn’t Happening

I didn’t hear of the Lindsay Clancy case at the time of it occurring in 2023, but I have now found myself side eyeing the ongoing trial.

On 24 January 2023, Clancy killed her three children: Cora, five; Dawson, three; and Callan, eight months. You might ask why there is still an ongoing trial when I can say so confidently that Lindsay Clancy killed her three children…she admits to doing it. That stipulated fact is not what her current trial is principally attempting to establish. Her defence isn’t claiming an unidentified intruder entered the house, or that her now ex-husband Patrick Clancy committed the acts instead. The dispute before the jury concerns criminal responsibility which has a specific meaning under Massachusetts law. A person lacks criminal responsibility where, because of a mental disease or defect, they lacked substantial capacity either to appreciate the criminality or wrongfulness of their conduct or to conform their conduct to the requirements of the law. The burden is not on Lindsay Clancy or her defence to prove that she lacked criminal responsibility, it is on the Commonwealth of Massachusetts to prove that she was criminally responsible beyond a reasonable doubt. The arguments and question the trial is really asking is “what was the condition of her mind when she killed her children?” which is difficult to answer.

The defence argues that Clancy was experiencing postpartum psychosis and was therefore not criminally responsible. They have called experts that have supported that interpretation and pointed to things like an over-prescription of medication and lack of medical care. Prosecution experts have disputed that she was acutely psychotic and argued that, although she was mentally ill, she retained the capacities required for criminal responsibility. That is the difficult case the jury have been hearing from inside the courtroom. Online, though, there is another happening and in that version Patrick Clancy killed the three children. Or he drugged Lindsay. Or he deliberately manufactured the conditions that caused the collapse. Or the person recorded on the CCTV while Patrick was collecting medication at CVS was actually a body double. I’m not exaggerating there, people have genuinely suggested a body double. The security camera footage, the phone records corroborating Patrick’s whereabouts, somehow just becomes another obstacle for the theory to route itself around rather than evidence against it.

Then there is “Forensic Astrology”. Once again, I am not joking. That is a term that some of these people are using for themselves. I’m torn on whether they don’t realise the clear oxymoron in that title, or whether they genuinely believe that astrology is in some way scientific. Someone call Baz Luhrmann and tell him Romeo and Juliet can be modernised once more:

“Feather of lead, bright smoke, cold fire, sick health……forensic astrologer”
- Romeo and Juliet, probably.

These absolute titans of deep investigative work can go on to TikTok, stick up their credentials which usually amount to how she felt a crystal vibrated in a certain way after farting with Mercury in retrograde, and spout off a load of nonsense and conclude with certainty that Patrick is actually the perpetrator. He is a Gemini, after all. And then the mainstream reporting around the trial are feeding in to these deluded claims by even bothering to give them air time or consideration instead of laughing them off with absolute derision. Maybe when Mercury stops being in retrograde over my anus the prosecution can call it to the stand and start to establish a chain of custody. What else can we do with the knowledge that people are producing crime charts, interpreting planetary positions and treating them as information capable of identifying hidden influences in crimes. It would just stay as funny but now that it has been given a perceived credibility by being featured on these law channels it has crossed over into publicly accusing a real person of murdering his children.

No evidence for it. Whereabouts during the killings corroborated through surveillance and phone records. Lindsay’s own defence not presenting another person let alone Patrick as the perpetrator. Yet the theories persist. And as you read deeper you come to realise that their theories appear increasingly resistant to falsification because contrary evidence is repeatedly incorporated as further evidence of the conspiracy. They are steadfast in their dream.

And this is where we return to our dreamer. A person sees something that feels wrong, they form a judgement, someone else agrees. Someone finds a discrepancy, interprets a facial expression, notices a timestamp. Someone else has a dream. An astrologer finds something in Gemini. Comments being theories; theories become the premise for more videos; the popularity of that video becomes evidence that lots of people have independently noticed the same thing. Eventually there is an entire system of mutually reinforcing judgements. An ouroboros not eating its tail, instead shoving its head up its own arse. Everything feels extraordinarily coherent and everything connects. Has it proven anything? Coherence inside a closed system proves very little about whether that system connects with the world outside it. It’s not created knowledge, it’s an organised dream.

Judgement Without Reason

You might be starting to feel uncomfortable, you may start to have a niggling question gnawing at the back of your brain. It’s OK, I’ll ask it.

Would you want these people on a jury of your peers?

Not people who enjoy true crime, we’ve seen the numbers it would disqualify an alarming percentage of the population. I mean someone who genuinely believes that the position of a planet can establish whether another human being committed murder. Imagine discovering after your own conviction that one of the twelve people deciding whether you would spend the rest of your life in prison was less interested in the forensic evidence than whether your parents were celebrating their anniversary or a birthday…I was also two-weeks early so is it based on the actual date and time of birth or when it should have been based on a full gestational period?

I don’t know about you, I would have concerns. It sounds medieval. If you went to the pharmacist because of a headache and they suggest you stick leeches on your eyeballs for two hours four times per week you’d probably at least seek a second opinion before diving in to worm contact lenses.

So what does a jury of our peers even mean? It should necessarily mean a jury composed of ordinary humans; each coming with prejudices, strange beliefs, emotional histories, political assumptions, bad memories and the occasional terrible night of sleep. The legal system knows this and doesn’t seek to resolve it by trying to source the non-existent perfectly rational 12. Instead it constructs procedures around imperfect people. Evidence rules about what can be considered, cross-examination challenging the testimony, judges provide instructions about the law, jurors deliberate with one another. The standard of beyond a reasonable doubt deliberately limit the confidence required before the state is permitted to impose the most serious of punishments. The jury is not expected to arrive to the trial equipped with pure reason, there is a process that is supposed to discipline judgement. But as I have already articulated earlier, there are a growing subset of these people that make it clear no amount of evidence will change their mind. That childish obstination is insidious and there is little to no expectation that they would consider the constructed procedures and would instead base their decision entirely on their dreams.

If you decide that, no you’d rather not have them on your jury, what is the alternative?

Would an AI be better at it?

Reason Without Judgement

Sounds almost dystopian doesn’t it, an AI jury? When it comes down to it, though, and given the case I have presented so far, how different is dystopian to medieval and which is preferable?

Sure at first glance the AI jury sounds seductively sensible. Give the system every item of admissible evidence, give it the full transcript, give it the relevant law and the judge’s instructions. Require that every conclusion identifies the evidence that supports it and make it distinguish direct evidence from inference. Make the uncertainty explicit. Isolate in a sandbox to prevent external influence from press coverage or social-media commentary or anything the court has rules inadmissible. The ever present human elements, the ones that the Judge asks the Jury to put aside, are naturally missing. No sympathy because somebody cried convincingly. No antipathy because someone didn’t cry enough. No dislike because a witness is perceived as arrogant or any impatience because the deliberations have gone on for three days already and everyone wants to go home and see their family. And most importantly, no astrology.

Except, that isn’t really what an AI system is. An AI jury would not be a magical box labelled OBJECTIVE REASON into which we feed evidence and receive justice. It would still be a designed system built around a model. There would still be human assumption and biases baked in to the training data, human written instructions, human model choice. Strings of human decisions made in the presentation of information and data, in how uncertainty is scored and articulated and what the system should do when two plausible explanations remain. Although the AI could potentially solve some human failures, models can fail in ways that humans do not. Over-weighted patterns that happen to correlate in training data, inventing details, behavioural changes under superficially small changes in framing are all just other failure modes. By removing emotion it doesn’t automatically produce objectivity; but it removes emotion from the decision making process.

I should be clear here, when an AI company describes a model as reasoning, that isn’t what Kant was referencing with reason. Kant’s reason is bound up with the attempt to organise judgements under principles and to constrain claims according to conditions that make knowledge possible. In AI systems “reasoning” is a functional description. The system can manipulate representations, compare claims, follow rules and produce chains of inference but there is no evidence that it cares whether justice has been served. That might sound like an advantage but look again and then you can realise that criminal judgement is not just a difficult classification problem; a jury doesn’t merely output GUILTY or NOT GUILTY. Twelve citizens are being entrusted with the authority to participate in the state’s judgement of another person. The state offloads the responsibility for what follows to those twelve.

Could AI still become useful within the process without allowing it to judge? It has potential to expose contradictions, trace provenance, test whether a conclusion actually follows from the evidence that has been presented. It could even show a jury where its reasoning has quietly acquired an assumption nobody has established. And maybe that is the distinction. Humans are quite capable of judgement without much reason. Machines may eventually become extremely capable of reason without judgement.

Neither outcome is particularly comforting on its own. It’s almost like asking to choose between living in Mega-City One where AI Judge Dredd enforces the law; or 16th Century England where looking at your neighbour in the wrong manner could get you accused of Witchcraft.

The Conversation We Should Be Having

I do have an opinion about the Clancy case but ultimately it doesn’t matter what I think. At the time of writing this, there is no verdict and I am neither her psychiatrist nor a member of the jury. I wasn’t there after the birth of her first child, second child, or third child and I have never observed a single second of the family as it was. Even by watching the broadcast aspects of the trial we are not party to all the information that the jury receive.

What we do know is that Cora, Dawson and Callan, three young children, will never get to experience the lives they should have had. That should probably be enough to resist turning their deaths into a puzzle requiring a hidden final twist. There is already something here worth talking about.

Postpartum psychosis is rare but serious and potentially extremely dangerous. It is suggested that 0.1% to 0.2% of women will be affected by it after childbirth, and underlying vulnerabilities such as Bipolar I can raise the risk incredibly with some figures suggesting 20%. Perinatal mental-health problems extend much further than psychosis and shame can make already frightening symptoms even harder to disclose. The narrative that is forming, though, from a lot of women that they were also on the verge of a psychotic break after childbirth and that this situation could happen to any of them is a dangerous conflation to make. There is a difference between the “baby blues”, postpartum depression, and postpartum psychosis. To start muddying the waters on that now risks then making the situation even worse than it is now. Just think, if there is no distinction being made between the “baby blues” and postpartum psychosis then how can an accurate risk assessment be made?

People need to know that intrusive thoughts, depression, anxiety and paranoia after pregnancy are things that can and should be spoken about without becoming an accusation about whether somebody is a good mother. It seems that they want to consider “Mother” as a monolith and ignore the vast differences between people. Families need to recognise the signs too, and the clinicians need the resources and systems to be able to respond properly when somebody asks for help. Whether that is the mother or the father who is highlighting a concern. There is an organisation trying to push that conversation forward.

The Heard Foundation works on perinatal mental-health awareness, stigma, access to care and the development of a dedicated treatment centre in Massachusetts. Their aim is to make it less likely that another family reaches the same catastrophic outcome.

Heard was co-founded by Patrick Clancy.