In 2020, a number of 5G masts were set on fire in the UK because of a belief that gained traction online linking 5G to the spread of COVID-19. Whoever committed this arson seems to have been motivated by the connecting of a series of “dots.” First (dot one), that 5G towers may weaken the immune system; second (dot two), that this made people more vulnerable to COVID; third (dot three), that 5G therefore, in a way, “caused” the pandemic.
It is easy to connect dots in ways that seem logical to us but do not withstand scrutiny. Sometimes these connections start with a fundamental misconception (such as that 5G weakens the immune system, which it does not), sometimes they start with correct, or at least plausible, information but go astray somewhere along the way. This was arguably the case with hydroxychloroquine during the COVID-19 pandemic. Early observational studies suggested there might be some benefit to its use. This was amplified rapidly through media and politics. But large RCTs, such as the RECOVERY trial, found no benefit. Nevertheless, hydroxychloroquine was touted for its supposed effectiveness, even as the data did not support this hype. The initial, plausible data point suggesting hydroxychloroquine may have some utility was quickly connected to other dots that had no bearing on its actual effectiveness, leading many to draw the wrong conclusions about the drug.
Closer to home, there has been much critique leveled at social science, in particular, for perhaps connecting too many dots, for making causal inferences too glibly. Are we maybe too quick to assert that a particular upstream neighborhood condition causes an individual health indicator? Are we connecting too many dots there? Is this, in part, why we have a replication crisis? We cannot avoid these questions if we are to shape a science worthy of the work we are trying to do, the world we are trying to create.
Why are we so quick to connect dots, even when these connections are suspect, the conclusions they suggest faulty? The answer lies, perhaps, in our fundamental nature. We look for patterns, and, when we do not find them, we are prone to imagining them. This is true, for example, when we look up at the stars and see constellations, animals, gods, heroes out of myth. Connecting these celestial dots, we make the sky a story. We need stories, and are quick to embrace a compelling one, even when it does not fully adhere to the facts, or even adhere to them at all. The work of science itself is, in large part, the work of making connections, of finding the associations between data points. The story science tells, at its best, is the story of truth sought and found. We connect the dots, and they can lead us to progress, technological breakthroughs, civilization itself. But we can connect too many dots too quickly, we can base these connections on assumptions and wishful thinking rather than fact, and this can get us into trouble.
It seems worth taking a moment, then, to consider the risk of connecting too many dots and asking how we can better engage with data points to ensure they lead us, and others, in the right directions.
We should perhaps first ask: what are some of the dangers of over-connecting dots? First, there is the danger of false certainty. When we connect the dots without taking time to ensure our connections are sound, our data points valid, we risk thinking our conclusions are more solid than they are. This can lead to mistakes, to the intellectual equivalent of confidently placing our foot where we assume there will be a step only to stumble when we find none there. False certainty can mean we categorize as fact what is still hypothesis and act like certain questions are resolved when they are still open. This runs counter to the core imperative, in science, of avoiding certainty as much as possible, of subjecting even our most confidently-held notions to scrutiny and review.
Second, there is the challenge of amplifying error. Insofar as a chain is as strong as its weakest link, when we build a chain of weak links, each point of weakness compounds the overall fragility of the whole. Ensuring these links are strong takes patient, thoughtful work, yet our present technological context disincentivizes such care. Online media is, among much else, a zone where faulty dot connecting thrives. It amplifies conclusions based on bad evidence, narratives that draw more from feeling than fact. When we fail to properly connect dots, the dynamics of algorithms and online media can ensure our mistake gets out into the world where it can do real harm.
Third, when we do not properly connect dots and then are found to have erred in this way, it can erode the public’s trust in what we say and do. During COVID, for example, there were occasions when we issued guidance based on connections that simply were not there or were not as clear as our confidence in stating them implied. This contributed to an erosion of the public’s trust in us, a deficit with which we are still reckoning. We owe it to the public to avoid saying or doing anything that reflects the kind of thinking that leads us to connect dots poorly. If, however, the demands of the moment force us to make recommendations in a context of incomplete and evolving information, it is on us to be clear about this uncertainty in our engagement with the public, to convey that we are doing our best with what we know and that what we know is incomplete rather than dots fully connected and associations fully clear.
Fourth, we should bear in mind, always, that when dots are not correctly connected, they can lead to policy missteps with serious real-world consequences. These can include interventions that cause active harm, or that simply waste time and resources that could be dedicated to more worthwhile initiatives. We can imagine, for example, the good that might have been done had all the time and attention given to hydroxychloroquine been used to promote more effective strategies for addressing COVID. A better pursuit of these strategies is one where we avoid the blind alleys in which we can find ourselves when we connect dots poorly in favor of approaches that improve health by reflecting better science.
Fifth, narratives based on incorrectly connected dots can fuel polarization, to the detriment of the public debate. When we are too quick to connect dots, when we neglect the patient, sometimes frustrating, work of slowly, methodically navigating from point A to point B of a given argument or hypothesis, we risk embracing oversimplified or inaccurate stories that can, when introduced into the public debate, harden partisan divides, close down conversation, or make intractable the open, data-informed conversations we should be having about issues of consequence for health. This was arguably the case in the summer of 2020, when many were quick to embrace certain narratives—on both sides of the issue—about race and policing, sometimes tuning out data-informed criticism that challenged the logic of these perspectives. This meant that the conversation generated much heat, but often not as much light as it should have, as participants in the debate retreated to their partisan corners.
How can we avoid connecting too many dots and the challenges that doing so can create? Three suggestions.
First, we can cultivate the intellectual discipline to resist connecting too many dots too quickly. We can ask questions like: is this inference actually true? Do these dots genuinely connect? This means pushing past how something may seem to find out how it is. An RCT, for example, may suggest a drug’s effectiveness in a certain narrow context, but perhaps replication will reveal it does not live up to its promise. It is on us to be slow and deliberate in this process, to take the steps that need to be taken to ensure what we are saying and recommending reflects the best thinking of which we are capable in the moment.
Second, we need to look beyond tropes and heuristics, to exercise the kind of judgement that lets us fully evaluate, and then communicate, the reality of a given intervention. This means recognizing that the tendency to connect dots poorly is a human one, that we are not exempt from it, yet we are in a position where, participating as we do in the work of science, we are expected to be less susceptible to it than others. Given this expectation, and the damage falling short of it can do to the public’s trust in us, we should remember that just because something “sounds true” does not mean that it is, and that just because something does not match our preferred narratives about the world does not mean it does not have validity. We should also take care to avoid blind reliance on methodologies which may be sound but only paint an incomplete picture. (This was well, and humorously, illustrated in a 2003 BMJ paper which conducted a systematic review of randomized controlled trials of “the effects of using a parachute during free fall,” and “were unable to identify any randomised controlled trials of parachute intervention.” The study concluded, “We think that everyone might benefit if the most radical protagonists of evidence-based medicine organised and participated in a double blind, randomised, placebo controlled, crossover trial of the parachute.”)
Third, we should set a high bar for ourselves and do our best, always, to clear it, maintaining an exacting standard for what counts as knowledge. Such a standard is bolstered by replication, triangulation across methods, and skepticism toward overly neat narratives. We should insist on replication across different contexts, attend to effect sizes, not just statistical significance, weigh the totality of evidence (systematic reviews, meta-analyses), and stay attentive to uncertainty and avoid premature closure when drawing our conclusions.
Connecting dots is natural, but when we over-connect them, we risk mistaking speculation for truth. The task is not to abandon pattern-seeking but to discipline it, to hold ourselves to a standard where connections are made with care so that our conclusions help build trust, and support health.
__ __ __
Also this week
ICYMI, I spoke with NPR about the administration’s health priorities and the importance of centering science in federal health policy.
"We see what we look for." - attributed to John Lubbock
"The times are urgent; let's slow down." - attributed to African saying
"One only sees what one knows." - attributed to Johann Wolfgang
Thank you for the clearer arc of view.
The passage shows that though pattern-seeking is natural, conclusions are only as strong as the evidence linking them, and over-connecting dots can produce false certainty, harmful narratives, and could result in unsustainable policy, I am learning a lot from THE HEALTHIEST GOLDFISH.