why you should hesitate over assisted dying

a philosophical long-read on the risks of legalisation

TLDR:

• The big question for the UK isn’t whether critics of the legalisation of assisted dying are relying on ‘slippery slope’ arguments. It’s whether the bottom of the slope is a relevant possibility

• The real risk lies not in the ‘slippery slope’ loosening of standard constraints on who can access assisted dying, but in the slipperiness of these supposedly tight constraints, themselves

• It’s difficult to predict when someone will die. It’s difficult to separate physical suffering from mental suffering. And it’s difficult to determine when consent-receiving retains relevance

• UK supporters of assisted dying should reconsider the conditions their support depends on, and remember the current institutional context of its would-be practitioners and overseers

---

I’ll admit upfront that I think the case against legalising assisted dying is over-determined. There are several separate arguments that convince me the UK should refrain from legalisation, and that assisted dying should be illegal everywhere else, too.¹ Rather than making those arguments head on, however, I’m going to push back against the seemingly second-order idea that the UK is immune to the ’slippery slope’ situations other places have experienced after legalisation.

I’m taking this approach largely because I don’t think slipperiness is as second-order as it might seem, and I’ll use some of my deepest concerns to explain why. But also because my primary aim here is to persuade people currently pushing for legalisation to hesitate and reconsider. And whilst I assume all these people are in favour of the tightly-constrained introduction of assisted dying, I assume few of them feel as positively about the situation at the bottom of the slope.

If you’ve been following UK discourse — as the current assisted dying bill makes its way through parliament — you’ll know that the ‘slope’ in question is the gradual expansion of access to assisted dying. And in particular, how places that initially enabled tightly-constrained access have, seemingly, systematically loosened their constraints.

At the extreme, imagine a place that enables access to assisted dying conditional on meeting the following three stringent requirements: 1) you’re an adult nearing death from a terminal illness, as confirmed by multiple doctors of relevant standing; 2) you’re suffering extreme pain, as confirmed by multiple doctors of relevant standing; and 3) you’ve persistently and reliably communicated your wish to die, for months or even years, as confirmed by multiple witnesses of relevant standing. Then, imagine access is gradually also made available to the following ‘expansion’ groups: 1) adults with non-terminal conditions; 2) adults whose suffering doesn’t consist in pain or any physical sensation, but various kinds of mental suffering; 3) adults who’ve communicated the wish to die only once or twice, or repeatedly but only for weeks or days; 4) adults whose suitability hasn’t been assessed and confirmed by multiple people of relevant standing; 5) children who meet the original requirements, or who track membership of one of the expansion groups.

I’m asking you to imagine all this, but it’s inspired by real-world instances. Consider Canada, where in 2016, it became legal for adults to access ‘medical assistance in dying’ (MAID), if they were suffering from a ‘grievous and irremediable medical condition’, and their ‘natural death’ was ‘reasonably foreseeable’. Since then, and following various access-focused court rulings, these constraints have been loosened: non-terminal conditions now deeming adults eligible for MAID include chronic pain and allergies, and the federal government has enquired into expanding access to under-18s.² Meanwhile, earlier this year, the Netherlands broadened its childhood eligibility requirements, making it legal to euthanise children between the ages of 1 and 12, having previously limited this to under-1s and over-12s. And in Australia, where access to ‘voluntary assisted dying’ (VAD) is determined at state level, the 2024 annual report of the Victorian government’s VAD board offers ‘reflections’ on how to ‘increase efficiency and reduce any unnecessary barriers to timely access’, including enabling doctors to initiate VAD discussions with patients.

Some UK MPs, however — most notably Kim Leadbeater, who introduced the current assisted dying bill to parliament — have claimed the UK is immune to such expansionism. The bill’s text hasn’t yet been published, but its full title refers to ‘allow[ing] adults who are terminally ill, subject to safeguards and protections, to request and be provided with assistance to end their own life’.³ And Leadbeater has stated, “Where there are countries where the law is broader, that was always how it started. So I think there is a perception around the slippery slope concept, which actually isn't reality”. In response, she’s received both praise and condemnation, and there’s been debate about the accuracy of her analysis.

Here, I’m going to agree and disagree with Leadbeater. I’ll begin by emphasising that you shouldn’t assume an argument is dependent on ‘slippery slope’ moves just because lots of people are saying so, particularly if they’re likely seeking rhetorical pay-off. Then, I’ll argue that the real risk to the UK lies not in the ‘slippery slope’ expansion of standard constraints on access to assisted dying, but in the slipperiness of these constraints, themselves. That is, I’m going to agree with Leadbeater that (at least at the level of potential) the expanded situation was indeed ‘always how it started’. But I’m going to disagree with her implication that the bottom of the slope isn’t, therefore, something that should worry supporters of UK legalisation.

Where is the slippery slope?

Let’s start with the slope. ‘Slippery’ implies sleight of hand, and when someone falls foul of the ‘slippery slope fallacy’, it’s because they’ve smuggled over-claiming into their argument. In its entry on fallacies, the Stanford Encyclopedia of Philosophy illustrates the slippery slope with the example of someone who decides to go to college because: ‘If you don’t go to college, you won’t get a degree; If you don’t get a degree, you won’t get a good job; If you don’t get a good job, you won’t be able to enjoy life; But you should be able to enjoy life; So, you should go to college’. The weakness of this argument, the author of the entry tells us, is in its ‘second and third causal claims’. And it’s easy to see that both of these ‘if-thens’ involve over-claiming: there are surely at least some good jobs available to people without degrees; there are surely at least some enjoyable ways to live that don’t involve having (got) a good job. So yes, when making formal arguments, we should avoid holding that X necessarily follows Y, when surely it doesn’t.

But this formal fallacy shouldn’t be confused with the sensible everyday practice of assessing the relative likelihoods of the differentially-valuable possible outcomes of potential actions. This everyday practice doesn’t entail assuming that possibilities are necessities: it’s just gaming stuff out in the awareness of relevant context. Sometimes, therefore, when Person A holds forth on a complex but worldly matter, and Person B shouts ‘slippery slope!’, it’s Person B who’s over-claiming. They’re committing the ‘fallacy fallacy’ and crying wolf. And sometimes this is because they’ve missed that Person A isn’t actually claiming that X necessarily follows Y, but that X is a relevant possibility. And sometimes it’s because they’ve missed that Person A is making an argument, rather than smuggling something in. But sometimes, Person B is simply seeking a rhetorical pay-off, at moral and logical cost.

The big question for the UK, however, isn’t whether critics of the legalisation of assisted dying are relying on ‘slippery slope’ arguments. Rather, it’s whether the bottom of the slope is a relevant possibility. In this context, I’m now going to argue that Leadbeater is in some sense correct that, in other places, the bottom of the slope was indeed ‘always how it started’. And that this should strongly influence supporters of UK legalisation to hesitate and reconsider.

Contestation over basic terminology

I’ll begin by clarifying what I mean by the contested terms ‘suicide’, ‘assisted dying’, and ‘euthanasia’. I take all three to refer to physical actions (or events centring on physical actions) successfully directed at purposefully causing the death of a person. And I take the core differences between the three to hinge on who does the ‘directing’, in that: 1) suicide is self-directed, 2) euthanasia is other-directed, and 3) assisted dying is both self-directed and other-directed. That is, all cases of both ‘assisted dying’, and ‘euthanasia’ involve the relevant physical actions of another person (or people) aside from the person whose death is the goal. But whilst in cases of ‘euthanasia’, it’s the other person alone who undertakes these actions (which is what happens in almost all instances of Canadian MAID), it’s always a joint enterprise in cases of ‘assisted dying’ (which is what the UK bill title proposes). To emphasise this distinction, I’ll refer to the latter as ‘assisted suicide’, rather than ‘assisted dying’.

Slipperiness arises, however, when we turn to my repeated reference to ‘physical actions’. I used ‘physical’ to distinguish between actions such as administering injections, and actions such as consent-giving. Of course, consent-giving does involve physical action (typically, uttering or writing), but it isn’t primarily physical. And consent-giving can sometimes count as ‘an action successfully directed at purposefully causing the death of a person’. But the point I wanted to make is that the kinds of actions central to determining whether a case counts as ‘suicide’, ‘assisted suicide’, or ‘euthanasia’ are primarily physical, whether they’re self-directed or other-directed.

Whilst this physical/non-physical distinction is useful at a simplified definitional level, however (and whilst with more words, we could tighten our three categories further), it’s hard to depend on this distinction to account for the nuances of a case in which, e.g., a paraplegic ends their life by sending an electronic command, via an implanted brain sensor, to a robot capable of administering a lethal injection. That is, it’s hard to work out which of the three categories this example neatly fits, if all we have is what I’ve given you. The complications involved in simply and neatly distinguishing between ‘suicide’, ‘assisted suicide’, and ‘euthanasia’, therefore, already signal slipperiness — and the risk of inherent expandability — at the level of basic terminology.

The supposedly tight constraints

I’m now going to turn to the slipperiness of the supposedly tight constraints. Remember the imaginary place where assisted dying is introduced on those three stringent requirements: 1) you’re an adult nearing death from a terminal illness, as confirmed by multiple doctors of relevant standing; 2) you’re suffering extreme pain, as confirmed by multiple doctors of relevant standing; and 3) you’ve persistently and reliably communicated your wish to die, for months or years, as confirmed by multiple witnesses of relevant standing. Surely, each of these seems the kind of requirement you’d put in place to ensure that assisted dying is used only to alleviate unbearable irrelievable suffering. And surely, most people in favour of legalisation are strongly, possibly solely, motivated by this compassionate concern. I’m going to argue, however, that the value of these requirements, to this limiting end, is under-cut by their slipperiness: that this slipperiness puts them at inherent risk of expansion.

The slipperiness of the first ‘stringent’ requirement — that you’re an adult nearing death from a terminal illness, as confirmed by multiple doctors of relevant standing — is clear, if you accept that, in almost all cases, it’s impossible to know when exactly a person will die, even if you’re the best-qualified doctor in the world. Indeed, another modern medical controversy hinges not on the difficulty of determining when someone will die, but on the difficulty of determining whether they already have. I’m thinking here of the contested notion of ‘brain death’, which is sometimes depended on to establish the permissibility of removing a person’s organs for transplant purposes.

But the difficulty of determining whether a possibly-dead person is still living pales in comparison to the difficulty of determining when exactly a definitely-living person will die. Moreover, whilst some of the problems involved in determining whether a possibly-dead person is still living are replicated in cases in which a definitely-living person has suffered the kind of catastrophic injury that makes death inevitable, there are many additional problems that arise in determining when someone will die who hasn’t suffered a catastrophic injury, but rather is suffering from a terminal illness.

Now, I’m aware that what I’m saying might seem overly reliant on the fact we can’t predict the future, and overly focused on pinpointing a moment rather than assuring inevitability. But the point I’m making isn’t that a doctor can never make a sufficiently reliable assessment that a person is sadly at, or nearing, the point of death. Rather, it’s that the uncertainties involved in making such assessments, in almost all cases, mean that a condition on which assisted dying is accessible only to those ‘nearing death from a terminal illness’ is unavoidably stretchy. Such a condition will, by its nature, capture people who were not nearing death from their terminal condition. And this isn’t just because we can’t predict natural death, and because we can’t check after intervention to see when natural death would have occurred. It’s also because assisted dying is explicitly aimed at bringing death forward. Therefore, this supposedly-stringent first requirement cannot be depended upon to maintain limited access to assisted dying.

Let’s turn to the second ‘stringent’ requirement: that you’re suffering extreme pain, as confirmed by multiple doctors of relevant standing. Now, without getting into a full debate about the meaning of ‘pain’, it shouldn’t be surprising that a measure aimed at relieving unbearable physical suffering might also be deemed useful for relieving unbearable mental suffering. After all, the two are often hard to separate. Partly because they can cause each other, but also because, even for a sufferer themself, it can be hard to know which is occurring.

Think about the phenomenon of feeling as if your brain is hurting. You know what I mean: you’ve been concentrating really hard on something really complicated for a really long time, and you feel more than mentally exhausted. There’s a real sense of pain! But it can’t reduce to the headache you’re also suffering. And the brain doesn’t have pain receptors. And it seems weird to think of it actually being your mind that’s hurting, not least because most people don’t buy the idea of the mind as separate from the brain. As it happens, I do, but if anything, this makes the ‘brain is hurting’ phenomena harder to parse. Perhaps, of course, it’s not actually pain you’re suffering in such an instance, or not any pain additional to the headache. But the slipperiness of the ‘brain hurting’ phenomenon points to why the supposedly stringent requirement of ‘unbearable physical suffering’ has some built-in ‘mental expansion’ available. And minimally, if it’s hard for you to know what kind of suffering you’re experiencing, then it’s infinitely harder for another person to confirm what it is.

Beyond this are intuitive knock-on concerns. I mean, if access to assisted dying is available for those who suffer physically, but not those who suffer mentally, then mightn’t that be unfair? Isn’t a well-known problem with most healthcare systems that they undervalue mental health? One way to respond would be to emphasise that mental suffering is not less important, but simply different, from physical suffering: that unbearable mental suffering can be treated with drugs and therapies. However, this just pushes the problem on to what counts as ‘unbearable’. Because, if the idea is that there’s a level of unbearable physical suffering that requires alleviation only in the form of assisted dying, then why isn’t there a similar level of mental suffering?

I’m not making this point to argue that either or both of these forms of suffering require assisted dying, but rather to emphasise the difficulty of keeping this second requirement ‘tight’. Indeed, my belief is that most people who claim they wish to die actually wish for something else: for their pain to be soothed; their anguish relieved; their guilt assuaged; even, in some cases, their boredom alleviated. Perhaps this sounds to you as if I’m failing to take the feelings of such people seriously enough. But if so, then I’d ask whether it’s you who’s failing in this way, by over-reducing these people’s feelings to ‘I want to die’. Moreover, if I’m correct here, then to help such people to die is to help them to do the wrong thing: you can’t depend on someone’s wishes to do something if it’s not what they wish! But I don’t need to conclude any of that. Rather, I just want to emphasise the in-built slipperiness of this second requirement.

Finally let’s turn to the third ‘stringent’ requirement: that you’ve persistently and reliably communicated your wish to die, for months or years, as confirmed by multiple witnesses of relevant standing. I think there are several slippery qualities to this requirement, but I’m going to focus on the following problem: how can we know that we hold adequate consent for something, when the consent in question can only be received in advance?

Imagine I’ve consented to you driving my car, saying, “Sure, you can use my car any time you like. No need to ask!” Imagine however that several months pass, and because I’ve been unexpectedly called away, we haven’t seen each other, and you can’t get hold of me. Imagine some other conditions change, too: your eyesight gets worse; you develop a bad drinking habit; the roads around my house deteriorate rapidly; some women in a nearby street take to shooting at all passing cars. If you require my consent to drive my car today, do you still have it? Next, imagine a highly intelligent man who, on being diagnosed with dementia, tells his family that when his condition becomes advanced, he wants them to help him to die. Over time, the man loses capacity for his previous intellectual pursuits, and it’s increasingly hard for him to communicate, but he finds new and immense joy in an unexpected grandchild.

In both these examples, it’s not straightforward to know whether consent is retained. Does our necessary consent-holding persist when we don’t have reliable insight into the consent-giver’s response to new relevant conditions? Can it persist when it’s not just that the consent-giver is temporarily unobtainable, but their capacity to consent is permanently reducing? These questions pose problems for the way in which the third ‘stringent’ requirement is predicated on the necessity of adequate consent-holding.

As it happens, my (somewhat controversial) view is that when consent-receiving retains moral relevance, it’s conditional on the consent-giver being able to retract it.⁴ But I don’t need to conclude that here, or that these consent complications make assisted dying impermissible. Rather, I’m simply emphasising some general risks in depending on the ongoing moral relevance of previous consent-receiving. And the particular risk of depending on it in cases where we know consent-givers are losing capacity. I’ll finish by noting, therefore, that this is complicated further by most instances of assisted dying involving some lack of consciousness for the dying person, before death.

Perhaps, however, you think I’m falling into the classic classical-liberal trap of over-focusing on consent.⁵ Perhaps you’ve noticed that the third requirement didn’t refer to consent, but rather to someone ‘communicating their wish to die’. If so, then answer this: is ‘assisted suicide’ possible without the consent of the person whose death is its goal? Does such a thing count as ‘assisted suicide’? And do you really want to argue that euthanasia is permissible without consent? Perhaps. But if so, you’re a long way from what’s being proposed by the current UK bill. And I’ll take this opportunity to reiterate that almost all instances of Canadian MAID are instances of physician-administered MAID: indeed, that self-administered MAID is illegal in Quebec.⁶

Some further conditionality

Finally, remember that the imaginary place didn’t impose its three requirements as alternatives: access to assisted dying was conditional on their combination. Therefore, each slippery strand slipperly intertwines. And that’s before we consider the particular context the would-be practitioners and overseers of UK assisted dying find themselves in. That is, whilst we can assume that people in the imaginary place suffer the same epistemic limits as us, nonetheless they find themselves in preferential institutional conditions. They’re not short of resources. They aren’t subject to non-transparent political rule. They don’t find themselves surrounded by scandals.

Perhaps your answer to why women in the UK receive low-quality maternity care, why the elderly struggle to access a GP,⁷ and why MPs don’t solve these important problems, is that resources are tight. If that’s your answer, however, then make sure you account for it before burdening doctors with ending the lives of their patients, and before burdening MPs with deciding whether doctors should do so. Again, this isn’t an argument against assisted dying. Rather, it’s a request to reconsider the conditions on which your support depends. Perhaps there’ll come a time when the resource problems, and the trust problems, disappear. Perhaps there’ll come a time when we know more, and we can reconfigure our thinking, beautifully sharp and devoid of slipperiness. Do you really believe we’re there, now?

---

A note about legality

You might’ve noticed that whilst this piece is about legalisation, I haven’t really talked about law. And you might be thinking, for instance, that one reason the UK shouldn’t fear the Canadian ‘slippery slope’ is that Canada’s expansion has been court-ruling-led, and the UK has a different legal system. The argument I’m making in this piece doesn’t require me to talk about the details of law and legal systems, however. That is — although an implication of my ‘slippery constraints’ conclusion is that law relying on such constraints is particularly open to interpretative contestation — I’m arguing that the primary risk of inbuilt-expansion lies at the conceptual level.

A note about democracy

It’s sometimes claimed that because there’s persistent popular support in the UK for the legalisation of assisted dying (it typically polls at almost 70 per cent), that it’s an affront to democracy to oppose legalisation. That doing so is anti-democratic! But surely, what counts as ‘democratic’, and its opposite, must involve more than the reflection of aggregated preferences. Otherwise, we could call the Murder For Fun company ‘democratic’, just because its employees collectively choose who to kill, and how to kill them. Therefore, beyond involving procedures for collective decision-making, there are some substantive preconditions that must be met for something to count as ‘democratic’. There are big questions about what these preconditions consist in, but it seems clear they must involve adequate respect for the equality of basic moral status: that is, the idea that the lives of all human beings are equally morally valuable. It’s not just that we shouldn’t call the murder company ‘democratic’, therefore, we should also refrain from referring to countries as ‘democracies’ if their approach to political participation violates this principle of equal moral status. That is, any ‘democratic’ nation that e.g., repeals women’s voting rights should be taken to be acting undemocratically, even if such a move matches the genuine preferences of every person living there, including every woman. Taken to its full extent, this means we should question the point at which ‘classic’ democracies, such as the UK and America, became democratic, if indeed they yet have. For now, however, we can conclude that, whilst ‘persistent popular support’ can be relevant to determining which legislative changes the members of a democracy should deliberate, it can’t suffice to determine the correct outcomes of such deliberation. Otherwise, we should confine the term ‘democratic’ to mathematical exercises, and forget about using it for political purposes.

1. My own starting point on this, as I might write about here sometime, is recognising there are certain things we cannot permit other people to do to our bodies. But I’m going to take a different approach in this piece, mainly because my starting point is underpinned by a particular self-ownership premise that even many fellow classical liberals won’t buy. And I’m especially keen to remain as ecumenical as possible today.

2. For insightful analysis of the Canadian situation, check out the writing and Twitter account of my friend Yuan Yi Zhu.

3. https://commonslibrary.parliament.uk/research-briefings/cbp-10123/

4. I recently wrote a paper about consent in which I make this argument, en route to trying to rescue social contract theory. It’s not published yet, but if you’d like to read it, I’d be glad to send you it.

5. ibid.

6. “Self-administration of MAID is permitted in all jurisdictions in Canada, except for Québec. There were fewer than seven deaths from self-administered MAID in 2022 across Canada, a trend consistent with previous years”: https://www.canada.ca/en/health-canada/services/publications/health-system-services/annual-report-medical-assistance-dying-2022.html#a5

7. Obviously there’s no need to hyperlink this.