In my earlier overview of the core claims of tobacco harm reduction (THR) advocates, I investigated three empirical claims in detail: that nicotine is relatively safe, that noncombustible tobacco and nicotine products help people quit smoking, and that misinformation is rampant about them.

THR also makes normative claims. This post focuses on one of the most important ones, inherited from the broader harm reduction tradition: the assertion that people who use psychoactive substances should be treated as agents in their own lives rather than as targets of public health optimization. The National Harm Reduction Coalition phrases this approach as affirming “drug users themselves as the primary agents of reducing the harms of their drug use.”

Effective altruists are often skeptical of rights-based or values-first claims, preferring to ask what improves welfare in expectation. This post argues that autonomy deserves more weight even within that framework as a guardrail against paternalistic public health calculations that can be unusually vulnerable to motivated reasoning.

Harm reduction starts with agency

Harm reduction began with practice rather than with theory. Before the movement developed any explicit ethical commitments, some of its earliest advocates were acting on its central values, to the point of breaking laws deliberately when they saw these as compromising the well-being of users of psychoactive substances.

One of the earliest harm reduction practitioners was a Yale public health student named Jon Stuen-Parker. In the early 1980s, when he arrived in New Haven, HIV was spreading rapidly around the country, and Parker wanted to do something about it. One severely affected group were people who injected substances like heroin intravenously and often shared the needles they used to do so. He began a series of efforts to educate people who used heroin about the dangers of needle sharing. A few years later, he started an initiative to exchange used needles for clean ones at a location in New Haven and later a mobile van in a number of US states.

Stuen-Parker was arrested 27 times in seven different US states for illegal possession of drug paraphernalia as he traveled the country providing clean needles to people who used heroin. In 1988, a federal ban on funding needle exchanges was passed by the US legislature, but he and others continued their efforts for decades despite the risks. Sometimes they invited legal consequences deliberately to raise awareness of the issue, as when ACT UP members openly distributed clean needles on the Lower East Side in New York City, accepting arrest in order to force a legal test case, one they ultimately won when a judge ruled the necessity defense applied to their actions.

Advocates of another emblematic effort, pill testing, have also taken big risks on principle. Adam Auctor, the founder of a multinational harm reduction organization called the Bunk Police, began his career by dropping out of his last semester of university. He used leftover student loan money to reverse engineer testing kits used by the police to detect adulterated pills and distribute them at music festivals where attendees have no other way of being confident what substances they’re taking. The organization has faced bans from promoters, financial losses from confiscations during raids of their tents, and threats of prison time from prosecution under the RAVE Act.

What motivates people to risk their livelihood and freedom in this way? Stuen-Parker and Auctor weren’t theorizing about public health ethics, they just saw a problem up close and took what they felt were the most practical steps to address it. But harm reduction, as it evolved from a loosely connected set of pragmatic practices to a coherent movement and philosophy, also started to develop an underlying moral stance. A key element of this stance is the emphasis on autonomy.

The phrase "any positive change," attributed to early harm reductionist Dan Bigg, expresses the idea succinctly. The person using the substance, not the institution providing the service, defines what counts as improvement. More recent philosophical work has pushed this further, arguing that harm reduction is not primarily about mitigating consequences at all, but about treating people who use psychoactive substances as equal members of society whose choices and preferences deserve the same respect as anyone else's. This is the distinction between harm reduction and mere pragmatism. The former asserts that even when coercive measures appear favorable in a narrow cost-benefit analysis of health outcomes, they start from the wrong premise about whose judgment should govern a person's choices about their own body.

Autonomy as a model adjustment

This values-based stance feels at odds with the utilitarian instincts underlying a lot of EA global-health reasoning. Autonomy may matter for well-being, but it often shows up as one input in a cost-benefit calculation rather than as a strong prior against coercion. A number of paternalistic interventions are endorsed and funded by impact-aligned organizations. Autonomy concerns are often treated instrumentally as one of many model inputs.

A few illustrative examples: GiveWell gave out a $15 million grant in 2021 to RESET Alcohol to lobby for more restrictive alcohol policies in LMICs. The reasoning was on a transparent and straightforward cost-effectiveness basis, with little discussion of autonomy concerns around policies designed to push people away from a form of consumption many adults value, even when it carries real risks and externalities. The section in GiveWell’s report on downsides includes “lost enjoyment from alcohol consumption” but the modeling applies what they themselves describe as a “fairly subjective” 10% downward adjustment in cost-effectiveness to account for this. The section on evidence-based solutions for the burden of alcohol-related health harms makes no mention of harm reduction approaches like Alcarelle, kava, supervised consumption, or the Sinclair method.

Coefficient Giving (formerly Open Philanthropy) explored tobacco control in detail in its 2023 shallow investigation. While devoting some space to harm reduction via e-cigarettes and a brief “side note” on snus, there’s no discussion of autonomy as a concern or tradeoff, as the discussion is restricted to cost-effectiveness on a dollar per DALY basis. The report’s author explicitly states that they “completely ignored any benefits smokers get from smoking” when developing their modeling on the effects of tobacco taxation and other programs intended to incentivize people to smoke less.

AIM CE has also recommended work on sodium reduction, including reformulation and sodium taxation, based on a 2023 deep report. The report included some exploration of negative impacts on well-being from less tasty food, but on a utilitarian basis rather than on a principle of autonomy basis. A dedicated section attempts to quantify the disvalue of reduced freedom of choice from a sodium tax, concluding it represents only an 8% reduction in the intervention's cost-effectiveness. The reasoning has a circular quality. The survey was administered to EAs, who were then noted to be disproportionately libertarian due to their high education levels, and their autonomy concerns are discounted on those grounds. The framework ends up treating the strength of someone's concern about autonomy as a reason to give that concern less weight.

The implicit rule in a lot of EA global health analysis is: do what works, without elevating autonomy as a privileged value per se, unless autonomy-violating interventions also impact well-being in a quantifiable way.

The galaxy brain problem in public health

So why should EAs or anyone interested in making the best possible impact with their time and money care about the harm reductionists’ rights-driven approach? If there are specific interventions that don’t have great evidence behind them, isn’t it enough to make an argument about that in cost-benefit terms, without having to engage in abstract philosophizing about utilitarianism versus deontology?

A major problem is that a cost-benefit analysis is only as good as the assumptions used to anchor it, and the more unconstrained those are, the easier it is to go astray and fall into rationalizing what we already believed for other reasons. Holden Karnofsky has argued that EA's maximization instinct creates constant temptations to rationalize, and that its track record owes more to the good judgment of the specific people involved than to anything in the framework itself. Vitalik Buterin makes a similar point, that act consequentialism suffers from what he calls “low galaxy brain resistance.” If your reasoning can justify any conclusion by constructing a sufficiently compelling calculation, it can take you to epistemically dangerous ground quite easily. One of Vitalik's specific examples of a lower-stakes but pervasive failure mode is "bad excuses for banning things for personal aesthetic reasons."

That failure mode is particularly acute in public health paternalism. Interventions targeting behaviors like drinking alcohol, smoking, and eating an unhealthy diet, address activities that many researchers and funders don’t engage in much themselves, or actively disapprove of. This makes it easy to underweight the preferences of the people whose choices are being restricted, not through bad faith but through unreflective cultural assumption. The cost-benefit framework doesn't generate its conclusions from neutral ground. It reflects the priors and sympathies of the people constructing it, and when those sympathies happen to align with the intervention being evaluated, the analysis is at greater risk of finding what it was looking for.

The advocates and funders of each of the interventions mentioned above are intellectually honest and self-aware enough to document and analyze the issues with coercive measures. The lost enjoyment from moderate drinking, the social and cognitive benefits of smoking, and the taste of salty food all show up in the analyses. But they receive remarkably little attention relative to the strength of the cost-effectiveness case being made, and in no case do they prompt comparable engagement with whether non-coercive alternatives might be preferable.

The case for a strong autonomy prior

This is where harm reductionist thinking can sharpen EA's epistemics in a specific and actionable way. While EA has its roots in philosophy that its early advocates aimed to grow into practical ideas for effective action, the arc of harm reduction’s history developed in the opposite direction. Starting from direct experience of what happens when public health authorities override the preferences of people they don't fully understand, its advocates then attempted to integrate these learnings into a coherent framework. Researchers who have tried to systematize harm reduction philosophy have identified autonomy as one of its core constitutive ideas because the movement learned through practice that good outcomes depend on it. The National Harm Reduction Coalition has codified this into a set of principles that treat the rejection of coercion and judgment over another person’s choices as foundational rather than as one consideration among many to be weighed.

This is essentially a rule utilitarian argument. The rule against paternalism doesn't need to be right in every individual case. It needs to be resistant to the motivated reasoning and cultural assumptions that make case-by-case consequentialist analysis unreliable in this domain. The case for that rule is stronger because harm reduction has repeatedly produced the kinds of results that paternalistic approaches often failed to deliver. Needle exchange made the riskiest version of IV substance use less dangerous, and the evidence now suggests that such programs cost-effectively reduce HIV transmission. Sweden’s unusually low smoking rates are hard to explain without the availability of snus as an acceptable substitute for cigarettes, and Japan saw cigarette sales fall dramatically after heated tobacco products became widely available. In each case, the less paternalistic intervention did not merely respect autonomy as a side constraint. The autonomy-respecting approach worked not despite meeting people where they were, but because it met people where they were.

EA already accepts this logic in other contexts. You don't run a fresh cost-benefit calculation every time someone proposes an exception to rules against dishonesty or manipulation, because you know the reasoning is manipulable. The harm reduction community's track record suggests that autonomy in health behavior deserves similar treatment.

An EA analyst who had internalized the autonomy prior would have approached each of the interventions above differently. That doesn’t mean they would necessarily reach different conclusions in each case, but they would ask harder questions earlier, take the preferences of the people being restricted more seriously, and take a more detailed look at alternatives that don't require overriding those preferences. In the three cases discussed above, there are well-evidenced autonomy-respecting alternatives that either go unmentioned or receive brief acknowledgment before being set aside without the scrutiny they would warrant in an analysis genuinely trying to find the most effective way to improve welfare. The harm reduction tradition has been asking these questions for decades. EA global health would benefit from engaging with what it found.