The Illusion of Proof: Confirmation bias and the Lucy Letby Case

Conspiracy theorists fall victim to a fundamental misunderstanding of how to evaluate and synthesise evidence. It’s not that they believe despite a lack of evidence, but rather that they fail to recognise how the evidence they possess is insufficient to support their improbable claims about the world.

While conspiracy theorists actively seek out evidence, their first error is confirmation bias – the tendency to search for information that supports their theory, while disregarding or swiftly dismissing counterarguments. They accumulate a collection of facts that align with their beliefs, relying on the sheer volume of evidence to convince themselves that they are correct. However, many of these individual pieces of evidence are often ambiguous, indirect, or inconclusive. Despite this, the cumulative effect creates the illusion of overwhelming proof in favour of the conspiracy.

In addition to confirmation bias, conspiracy theorists often fall prey to other cognitive distortions. One such example is anchoring bias, where they fixate on their initial beliefs and find it difficult to adjust these, even when confronted with new or contradictory evidence. Similarly, the Dunning-Kruger effect comes into play, where individuals overestimate their competence in evaluating complex information, leading them to draw misguided conclusions with misplaced confidence.

Another reason why conspiracy theories persist is their tendency to offer simple explanations for complex phenomena. In a world full of uncertainty, it is far easier to believe in a single, coherent narrative than to grapple with the ambiguity and randomness that often characterise reality. These theories provide a sense of clarity and control, which can be comforting when faced with a chaotic world.

Vaccine denial is a prominent instance where individuals exhibit flawed approaches to evaluating evidence. Vaccine sceptics often start with a preconceived belief that vaccines are harmful or unnecessary. Driven by confirmation bias, they actively seek out information that supports this belief while disregarding a vast body of scientific evidence that contradicts it.

These individuals might focus on anecdotal reports of adverse reactions to vaccines, amplifying rare cases and presenting them as common occurrences. They often cite discredited studies or misinterpret legitimate research to suggest a link between vaccines and conditions like autism—despite extensive studies showing no such connection. The anchoring bias keeps them fixated on initial misinformation, making it challenging for new, contradictory evidence to alter their views.

The Dunning-Kruger effect further exacerbates the issue, as individuals with limited medical or scientific training overestimate their understanding of complex immunological data. They might mistrust experts and authoritative bodies, believing instead that they have uncovered truths that the wider scientific community is either oblivious to or is actively suppressing.

Moreover, vaccine sceptics frequently misinterpret statistical data. For instance, they may argue that because some vaccinated individuals still contract a disease, vaccines are ineffective—overlooking the concept of herd immunity and the fact that vaccines significantly reduce the overall incidence and severity of diseases.

This selective gathering and interpretation of evidence create an illusion of substantial proof against vaccination. The sheer volume of misinformation shared within like-minded communities reinforces their beliefs, making it increasingly difficult to penetrate this echo chamber with factual information.

The same type of confirmation bias that fuels vaccine and climate change denial can also play a devastating role in legal proceedings. In the case of Lucy Letby, a neonatal nurse accused of murdering multiple infants, it is possible that confirmation bias contributed to the formation of a narrative of guilt, despite ambiguities in the evidence. Investigators, overwhelmed by the sheer number of infant deaths under her care, may have sought out patterns that fit their initial suspicions, ignoring or dismissing evidence that pointed elsewhere. Much like in conspiracy thinking, the accumulation of ambiguous, indirect evidence can create the illusion of overwhelming proof, leading to serious miscarriages of justice.

In criminal trials, several safeguards are built into the process to prevent the kind of biased evidence interpretation seen in conspiracy theories. The role of the police and prosecution is to present the best evidence they have to demonstrate the accused’s guilt beyond reasonable doubt. Importantly, the system provides a defence team specifically tasked with challenging the prosecution’s case, exposing weaknesses and inconsistencies in the evidence. This adversarial structure is intended to ensure that the court evaluates all the evidence in a balanced way, avoiding the risk of confirmation bias.

However, if the defence fails to adequately challenge the prosecution’s case (as is being repeatedly suggested by the series of bizarre decisions the defence made not to call their own expert witnesses), or if crucial evidence is overlooked or misinterpreted, the system can break down. In such cases, confirmation bias can slip into the process, where ambiguous or incomplete evidence starts to seem overwhelming. Much like in conspiracy theories, where a few suggestive but weak pieces of evidence are amplified into a larger, misleading narrative, this can lead to a false conviction, with the accused painted as guilty despite significant gaps or uncertainties in the case.

---

So, what may have happened in the Lucy Letby case?

At the heart of this tragic situation could have been a fundamental misunderstanding of the most likely explanation for an apparent rise in infant mortality. Just like in a conspiracy theory, where unlikely scenarios are built up through selective interpretation of evidence, the prosecution’s case against Lucy Letby was designed to demonstrate an improbable conclusion: that an intensive care unit nurse had deliberately murdered multiple babies in her care. While this was the narrative presented, it rests on the highly unlikely assumption that a psychopathic mass-murderer had infiltrated a neonatal ward undetected.

However, in any rational assessment of evidence, the least likely explanation should be approached with the greatest caution. In this case, rather than immediately assuming sinister intent, other explanations should have been more thoroughly explored. One plausible alternative is that the increase in infant deaths could have been the result of chance. Random variations in mortality rates are not uncommon in high-risk medical settings, and a cluster of tragic but natural deaths could easily be misinterpreted as a pattern of deliberate harm.

Another explanation for the spike in infant deaths on the ward may lie in systemic issues within the hospital itself. During the trial, it was revealed that the neonatal unit was understaffed and under-resourced, a reality that is unfortunately common in many healthcare settings. There were also suggestions of experienced staff having been got rid of, and serious hygiene issues on the ward. With stretched resources, insufficient staffing, and potentially exhausted or overworked medical personnel, the quality of care may have suffered, inadvertently leading to poorer outcomes for the infants. Such conditions are known risk factors for increased mortality, and without addressing these issues, the assumption of malicious intent could obscure the true causes of these deaths.

Furthermore, the care quality issues on the unit were documented even before the tragic deaths. These could include a lack of proper training, delays in interventions, or equipment failures—any of which might lead to an increased risk of infant mortality. It is far more probable that these systemic failures, rather than the presence of a malevolent individual, contributed to the unfortunate rise in deaths.

If these more likely explanations were inadequately explored, the resulting narrative could resemble that of a conspiracy theory, where improbable assumptions—such as the presence of a serial killer in a medical environment—are made to fit the data, rather than considering more straightforward, if uncomfortable, truths.

In such complex situations, a Bayesian approach to evidence could have helped ensure a more balanced evaluation. Bayesian reasoning encourages starting with prior probabilities—in this case, the likelihood of a nurse in a neonatal unit being a mass murderer—and then updating those beliefs in light of new evidence. Given that the prior probability of a serial killer nurse is extraordinarily low, the evidence presented would need to be overwhelmingly strong to shift the likelihood in favour of such an explanation.

But as we have seen, the evidence against Lucy Letby included ambiguous medical data, witness testimonies from overburdened staff, and a narrative that seemed to fit the tragedy rather than clearly prove guilt. When considering alternative explanations like chance variation and systemic failures, the Bayesian framework would likely suggest that the mass-murderer hypothesis is far less probable than other, more straightforward causes for the deaths

---

This is not the first time that flawed reasoning and statistical misunderstanding have led to the conviction of an innocent nurse. The case of Lucia de Berk, a Dutch nurse wrongfully convicted of multiple murders in 2003, provides a striking parallel. Like Lucy Letby, Lucia was accused of deliberately harming patients after a suspicious pattern emerged—she was present during a number of medical incidents and patient deaths. However, as Ben Goldacre explained, the pattern that convicted her may well have been nothing more than a random cluster, misinterpreted as evidence of guilt.

All across the world, nurses are working on wards where patients die, and it is inevitable that on one ward, in one hospital, in one town, in one country, somewhere in the world, you will find one nurse who seems to be on a lot when patients die. It’s very unlikely that one particular prespecified person will win the lottery but inevitable someone will win: we don’t suspect the winner of rigging the balls.

Goldacre illustrates this with the metaphor of the “Texas sharpshooter” phenomenon: imagine firing a thousand bullets at the side of a barn while blindfolded. After taking off the blindfold, you find a few bullets close together and paint a target around them, then declare yourself a sharpshooter. This is, of course, absurd—but it highlights a crucial flaw in how we interpret patterns. In the case of Lucia de Berk, a cluster of tragic deaths on her watch seemed highly suspicious, but it could easily have been a product of chance. Chance clusters happen routinely. Baby murdering nurses are very rare.

In Lucia’s case, this misunderstanding of random clustering led investigators to believe that the deaths were more than just an unfortunate coincidence. The courts failed to account for the fact that unusual patterns can and do arise purely by chance, especially in high-risk environments like intensive care units. It took years of legal battles and expert reviews before Lucia de Berk was fully exonerated, with the courts eventually recognising the role that confirmation bias and statistical errors played in her wrongful conviction.

This type of error has a name for it in law: The Prosecutors Fallacy.

The prosecutor’s fallacy is a common error in reasoning, particularly in legal contexts, where the likelihood of evidence is misinterpreted in a way that favours the prosecution’s case.

In essence, the fallacy occurs when the probability of observing certain evidence if the defendant is innocent (i.e., the likelihood of the evidence under the null hypothesis) is confused with the probability that the defendant is innocent given the evidence. These two probabilities are not the same, but they are often mistakenly treated as such.

To restate Goldacre’s analogy, the probability that you will win the lottery is not the same as the probability that someone will win the lottery. The probability that some nurse somewhere will have a high death rate against her shifts is not the same as the probability that any particular nurse, such as Letby, will see a high death rate. Just as we should not accuse Lottery winners as being cheats we should not accuse nurses unfortunate to see lots of deaths as being serial killers. This can appear to be counterintuitive, and can be difficult to understand, but courts need to be aware of it. (Good explainer here.)

This cautionary tale should have served as a warning, yet it seems similar errors may have occurred in the case of Lucy Letby. Just as with Lucia de Berk, the suspicion surrounding Letby seems to have arisen from a statistical anomaly: an unusual number of infant deaths occurring during her shifts. However, once the hospital and police anchored onto the idea that a nurse was responsible for these tragedies, all subsequent evidence was interpreted through that lens. Instead of considering more probable explanations—such as chance variation or systemic failures in the unit—the investigation was shaped by this initial assumption. Every piece of ambiguous evidence was scrutinised to support the theory of a mass-murdering nurse, much like painting a target around clustered bullet holes. This anchoring bias, combined with the failure to account for random clustering, may have transformed a series of tragic but natural deaths into something far more sinister in the minds of the investigators.

---

The investigation into Lucy Letby’s alleged involvement in the infant deaths on her ward appears to have taken an unlikely scenario—a psychopathic nurse deliberately harming babies—as its leading hypothesis from the start. At the outset, there was no solid evidence to suggest foul play beyond the fact that Letby had been present during a cluster of deaths. Letby herself exhibited none of the behavioural or psychological traits commonly associated with serial killers. In fact, she was widely regarded as a compassionate and competent nurse by her colleagues and patients’ families. Furthermore, initial medical assessments, including those made by coroners, had concluded that the infant deaths were due to natural causes. Yet, despite these indicators, investigators latched onto the idea that a nurse was responsible, allowing this unlikely explanation to frame the entire investigation moving forward.

From a Bayesian perspective, the hypothesis that Lucy Letby was a psychopathic nurse intentionally murdering babies is extraordinarily unlikely. Bayesian reasoning starts by considering the prior probability—the likelihood of a hypothesis before any evidence is introduced. In this case, the idea that a nurse with no prior history of psychological instability, no behavioural red flags, and a reputation as a diligent and caring professional could be responsible for a series of murders would carry an astronomically low prior probability, perhaps on the order of one in a million.

In Bayesian terms, this would mean that the bar for evidence would have to be set extraordinarily high. To shift the probability toward the conclusion that Letby was guilty, the prosecution would need to present extraordinary evidence—evidence so compelling that it would overcome the implausibility of the hypothesis. The reason for this is simple: when we begin with a hypothesis that is highly improbable, only extremely strong, direct, and unambiguous evidence should push us toward believing that it is true.

---

So, was the court presented with “extremely strong, direct, and unambiguous evidence” that Lucy Letby was responsible?

The answer is increasingly appearing to be no. Rather than uncovering overwhelming and clear-cut evidence, the case against Lucy Letby seems to have evolved into a fishing expedition. Investigators and prosecutors, having anchored onto the highly improbable hypothesis that Letby was a psychopathic nurse responsible for multiple infant deaths, began combing through every detail of her work, searching for signs that could link her to the tragic events on the ward.

What followed was an extensive re-examination of medical cases in a retrospective attempt to find evidence of foul play. Autopsies were redone, medical records were scrutinised, and expert opinions were sought, not to objectively determine the cause of death, but to confirm the hypothesis that had already taken root.

One significant example of this can be seen in the case of air embolisms, which became a focal point in the trial. Dr. Dewi Evans, a medical expert for the prosecution, suggested that the presence of air bubbles in some of the babies’ bloodstreams could be the result of deliberate injections. However, this interpretation was far from definitive. When asked to assess the medical findings, Dr. Arthurs, a radiology expert, told the court that the appearance of air bubbles was “consistent with, but not diagnostic, of air having been administered.” In other words, while the finding was consistent with the hypothesis of foul play, it did not prove it. There could be many other explanations for the X-Ray observation beyond deliberate injection of air into the baby.

This distinction is critical. Just because evidence is consistent with a hypothesis does not mean that the hypothesis is true. This is a common error in reasoning, similar to the way conspiracy theorists latch onto pieces of evidence that fit their theory while overlooking alternative explanations. In the Letby case, the fact that air bubbles could have been caused by natural processes or medical interventions was seemingly downplayed once the narrative of intentional harm had taken hold. Rather than providing “extraordinary evidence” to support such an extraordinary claim, this line of reasoning merely offered a possibility, whilst other more probably explanations were discounted.

A similar dynamic occurred with other unusual symptoms and medical findings in the babies. For example, some infants showed discolouration of the skin, which the prosecution argued could be a sign of deliberate harm, such as smothering. However, this symptom can also occur as a natural part of the body’s shutdown process in critically ill infants. Just like the air embolisms, the discolouration was consistent with foul play but far from definitive evidence of it.

This approach—where any unusual sign or symptom was interpreted in light of the preconceived notion of Letby’s guilt—mirrors the flawed thinking found in conspiracy theories. In conspiracy circles, it’s common to amass pieces of ambiguous or circumstantial evidence that are “consistent with” the theory, without stopping to consider whether they actually prove it. In the Letby case, once the narrative of guilt had taken root, evidence was sought and interpreted through that lens, even if alternative, more benign explanations were just as plausible, or even more likely.

It’s important to note that many more babies tragically died in the neonatal unit than were included in the case against Lucy Letby. The narrative that Letby was present at “all” of the deaths considered in the trial is technically true, but only in a trivial sense. In reality, investigators focused selectively on a subset of deaths that seemed suspicious and could be linked to Letby’s presence, while ignoring other cases where she was also working but no foul play was suspected. The fact that Letby was present at some of the deaths doesn’t provide proof of her involvement—it’s simply the result of her routine presence in the high-risk unit. Had the investigation considered the full picture, including all infant deaths during Letby’s time at the hospital, it may have become clear that her presence during these tragedies was not statistically significant or suspicious in itself.

This selective reasoning is yet another example of the Texas sharpshooter fallacy. Investigators essentially “drew a target” around the cases where Letby was present, ignoring the broader pattern of infant deaths on the unit. By focusing on a small, specific cluster of incidents that could be tied to Letby, they created the illusion of a strong connection between her presence and the deaths. But, just as with the Texas sharpshooter analogy—where the marksman draws a target around a few clustered bullet holes after firing at random—this pattern is artificial. In a busy neonatal unit with many fragile infants, deaths are unfortunately not uncommon, and it is inevitable that a nurse on regular shifts would be present during some of them.

The fishing exercise extended beyond medical evidence to Letby’s personal life. Investigators examined her personal diaries and social media messages, looking for anything that could hint at guilt or motive. This included the discovery of a note in which Letby had written “I am evil, I did this.” The prosecution seized on this as a form of confession, despite the fact that the context suggested it was more likely an expression of her distress under mounting accusations, rather than an admission of wrongdoing. The note also said “WHY ME? I haven’t done anything wrong”. The most likely explanation is this note is the expression of a young women in mental turmoil and breakdown under the pressures of investigation.

This follows the same flawed reasoning: the note was consistent with guilt, but far from definitive proof. It is also consistent with her being innocent. When viewed through the lens of confirmation bias, any evidence—even ambiguous or contradictory—can be interpreted to fit the narrative. The investigators, having already anchored on the theory that Letby was a serial killer, interpreted this emotional expression as damning, rather than considering the stress and pressure Letby must have been under as the investigation unfolded.

In both conspiracy thinking and this investigation, the problem arises when evidence that is consistent with a hypothesis is mistaken for definitive proof, leading to the entrenchment of a potentially false narrative. In Letby’s case, the fishing expedition failed to produce the extraordinary evidence needed to support such an extraordinary claim. Instead, weak, indirect and ambiguous findings were reinterpreted to fit a preconceived conclusion, overlooking more likely explanations for the tragic deaths.

---

Conspiracy theories thrive on offering simple explanations for complex and often disturbing events. In the case of the Lucy Letby trial, the idea that a single, psychopathic nurse could be responsible for a string of infant deaths provides a more straightforward, albeit chilling, narrative than confronting the possibility of systemic failure. The reality of a severely understaffed and under-resourced neonatal ward, where overworked staff struggle to provide optimal care, is a far more complex and politically fraught issue to address.

The National Health Service, like many large, publicly funded institutions, is not immune to systemic shortcomings. Staffing shortages, limited resources, and increasing pressure on healthcare workers are well-documented problems, and they create an environment where errors, oversights, and even tragic outcomes become more likely. But accepting that babies died because of these complex, structural issues forces us to confront difficult political and societal questions: Is the NHS adequately funded? Are our healthcare systems capable of safeguarding the most vulnerable, even in times of strain? Do these deaths reflect broader failings in the way we manage public health?

In contrast, the narrative of a “baby-killing nurse” offers a simpler, more emotionally satisfying explanation. The idea of an individual, acting with malice, makes the deaths feel less random, less the result of an uncaring or overburdened system. It allows the public to direct their outrage and grief toward a single person rather than grapple with the uncomfortable idea that preventable deaths might be the product of bureaucratic and political failings. It’s easier to believe in the existence of a psychopath than to reckon with the reality that babies may have died because of inadequate staffing, lack of resources, or flawed management decisions.

For the parents of the infants who died, this shift in narrative had a profound emotional impact. Initially, these families endured the unimaginable grief of losing a child, an experience already filled with pain, guilt, and questions. For many, the original conclusions from coroners and medical staff—that their babies had passed due to natural causes—may have offered some comfort or at least a sense of closure, however incomplete.

But when the investigation turned toward Lucy Letby, these same parents were forced to relive their grief, now compounded by the horror of believing that a monstrous human being had murdered their children. The trauma of losing a child in a medical setting—a place where they should have been safe and cared for—was amplified by the idea that their trust had been betrayed in the most horrifying way possible. Parents who were once grieving natural causes were suddenly facing the chilling possibility that their children had been deliberately harmed by someone they had trusted.

This shift in narrative, from natural causes to foul play, may have offered a simpler explanation, but it also reshaped the parents’ grief into something even darker. The focus on Letby as a single source of evil allowed a contained explanation for the deaths, but it prevented families—and the public—from considering the far more difficult reality that these tragedies may have been the result of systemic issues within the NHS. Acknowledging that their babies died due to a failure of the healthcare system—rather than malevolence—would raise deeply uncomfortable questions about how and why such failures were allowed to happen.

In a way, the Letby case reflects a broader societal tendency to engage in conspiratorial thinking when faced with overwhelming complexity. Whether it’s attributing political unrest to shadowy elite groups or assigning blame for tragic medical outcomes to a single person, the desire for clear, simple explanations is deeply embedded in human psychology. In both cases, this thinking offers a false sense of control—by identifying a villain, we convince ourselves that we understand the problem and can therefore fix it.

But the reality is often messier. If we accept that these deaths were the result of systemic failures, then the solutions become much more complicated. Improving staffing levels, addressing underfunding in the NHS, and fixing quality control issues require time, resources, and political will. These are much harder problems to solve than simply convicting an individual and assuming the threat is over.

Beyond the emotional impact on the parents, the Lucy Letby case also raises serious concerns about the reliability of the court system in high-stakes, complex trials. Our society depends on the justice system to be fair, efficient, and accurate in its pursuit of truth. When the courts get it wrong—particularly in cases with such devastating consequences—it sends a chilling message to all of us. If our courts can be misled into convicting someone based on weak or misinterpreted evidence, it should make us question the safeguards we rely on to prevent such miscarriages of justice.

The reliance on expert witnesses in the Letby trial, for example, demonstrates the critical role that interpretation plays in these cases. But experts, too, are human, and they are susceptible to the same biases as the rest of us. Once the hypothesis of a “baby-killing nurse” took hold, the prosecution’s expert testimony became shaped by a narrative that was difficult to challenge. This case, like others before it, highlights the Prosecutor’s Fallacy, where statistical anomalies and ambiguous evidence are presented as though they definitively point to guilt, when in reality they do not.

It would be easy, in the wake of a case like Letby’s, to demand simple solutions. One might suggest implementing tick-box checklists for spotting “killer nurses” or introducing stringent screening processes for healthcare workers to catch those who might do harm. After all, wouldn’t it be more comfortable to think that evil is obvious and detectable with a few more rules in place? Yet, these simplistic solutions ignore the deep complexity of both human behaviour and the systems that surround us.

Today the Thirwal Inquiry begins to supposedly investigate how “serial killer” Lucy Letby was able to work undetected at the Countess of Chester hospital during a series of tragic infant deaths It’s very scope seems to overlook the possibility of alternative explanations. The inquiry begins with the assumption that Letby is guilty, focusing only on how the hospital and the NHS system failed to prevent her actions, without critically re-examining whether she was responsible for the deaths in the first place. By doing so, it risks cementing a narrative of guilt that may not only be flawed but also prevents an honest exploration of the systemic factors—such as understaffing, resource shortages, or medical complications—that could have played a larger role in the deaths of these vulnerable babies.

The inquiry’s narrow scope—simply accepting Letby’s conviction—will likely result in superficial findings and recommendations. If it doesn’t address the possibility that systemic failures in the NHS neonatal unit were the root cause, any proposed “improvements” will be equally superficial. Rather than addressing deep-seated problems within the healthcare system, such as inadequate staffing, insufficient training, and overwhelming pressure on neonatal units, the inquiry risks producing symbolic reforms that fail to get to the heart of the issue. Worse still, it leaves the very real possibility that an innocent woman will remain imprisoned for the rest of her life, with the broader public accepting a dangerously simplified explanation for a highly complex tragedy.

The failure to scrutinise alternative causes for these infant deaths does a disservice to the families of the victims, who may never truly know what led to the deaths of their babies. Without examining the wider context—such as whether the NHS culture itself contributed to these tragedies—the inquiry risks focusing solely on (perhaps) non-existent procedural lapses, while ignoring the uncomfortable possibility that these deaths might have been the result of systemic NHS failures that were misinterpreted as foul play.

---