Suicidal empathy? When compassion replaces judgement

During 2025 and 2026 the phrase ”suicidal empathy” spread from an academic niche into ordinary political debate. Behind it is Gad Saad, professor of marketing at Concordia University in Montreal and formerly the holder of a research chair in evolutionary behavioural science, who argues in the book Suicidal Empathy: Dying to Be Kind that Western societies have allowed a fundamentally good trait — the capacity to feel with those who suffer — to grow so strong that it overrides the capacity to assess consequences.

The term is loaded, and deliberately so. It is not a diagnosis, not a measure and not a research field. It is a polemical metaphor, formulated to force a question: what happens in a society where the desire to be good becomes more important than testing whether the goodness works?

That question is reasonable, and it is old. The problem arises when the metaphor is treated as an explanation — when one moves from ”empathy can be partial” to ”empathy causes national decline” without anything in between having been measured. This analysis tests each link separately and states openly where the evidence runs out.

The conclusion in brief: the premises hold, but the application to Sweden is not empirically established. The distortions of empathy are well documented in psychological and political-science research. Swedish agencies have in several cases documented that control and follow-up failed. But no Swedish study connects the two — and that connection is precisely what the thesis requires.

A recurring error in debate is to treat empathy as a single trait one has more or less of. Research distinguishes at least three things: affective empathy (feeling what the other feels), cognitive empathy or perspective-taking (understanding the other's situation without necessarily sharing the feeling) and empathic concern (caring about the other's wellbeing and wanting to help).

The distinction is not academic. Jean Decety and Jason Cowell showed in a 2014 review that empathy and moral judgement rest partly on different processes, and that strong affective empathy can lead to bias as readily as to fair decisions. Someone who feels deeply for one person's suffering does not automatically become better at judging what is right for everyone involved — often the reverse.

Daniel Västfjäll, Paul Slovic and colleagues documented in a 2014 PLOS ONE study that the willingness to help actually declines when the number in need rises from one child to two. The phenomenon is called ”compassion fade”: feeling does not scale with need, it scales with recognisability. One face weighs more than a statistic, however many are hidden inside the statistic.

Elizabeth Simas, Scott Clifford and Justin Kirkland published a study in the American Political Science Review in 2020 showing something still more uncomfortable: higher empathic concern correlates with stronger animosity towards political opponents. Empathy thus functions not as a universal adhesive but as an in-group one — it is mobilised for one's own side's victims.

This is the research-solid core of empathy criticism, and it long predates Saad. Barbara Oakley and colleagues collected work on ”pathological altruism” in 2012, and Paul Bloom argued in Against Empathy (2016) for ”rational compassion” — care governed by principles rather than by who happens to be visible.

At the same time one thing should be said plainly: one of the most cited building blocks of empathy criticism, the identifiable-victim effect, has proved weaker than the debate suggests. Lee and Feeley's 2016 meta-analysis finds a small average effect with wide variation between studies. The direction holds; the magnitude is overstated.

Saad's argument has a clear structure, and it deserves accurate restatement before it is tested. He starts from an evolutionary premise: humans have an evolved disposition to protect whoever appears weak and vulnerable. In most environments that disposition has been constrained by cost — scarce resources, obvious risk, immediate feedback when help goes wrong.

In modern, affluent societies, Saad argues, those brakes have weakened. The cost of a failed act of goodness is rarely borne by the person who performed it, and it often becomes visible only much later. At the same time, concern has become a status marker: expressing it brings social reward, questioning it brings social punishment.

The combination is, for Saad, the dangerous part. When a position is morally coded it ceases to be tested empirically, because the testing itself is read as a moral transgression. Institutions then persist with measures that do not work, not out of stupidity but because the alternative would be to appear heartless.

It is a coherent hypothesis. It is also, as formulated, effectively untestable: it lacks outcome measures, it lacks a control group, and it lacks a definition of when a society would count as ”suicidal”. Any outcome can be read into the model after the fact, which is precisely what Popper warned was the mark of an unscientific theory.

It is also worth noting that several of the mechanisms Saad describes — status-driven moral signalling, costs borne by third parties, absent evaluation — are well known in institutional economics and public-administration research under entirely different names, and there without any need to posit empathy as the driver.

A concept like ”suicidal empathy” consists of several links, and they have very different evidence status. Conflating them is the most common error in the debate — among those who embrace the term and those who dismiss it alike.

The table below separates the links. The first three rows concern the psychology of empathy and have strong support. The fourth shows a building block that is often overstated. The fifth concerns a descriptive concept without a measurement instrument. The last concerns the conclusion itself — and there is nothing there.

If criticism of untested benevolence is to be constructive, it must be able to describe the alternative. The alternative is not less compassion. It is compassion combined with consequence-testing — two properties that vary independently of one another.

The quadrant below is an explanatory figure, not a measurement. The point is that there are two ways to go wrong, not one. Anyone who warns only about untested benevolence risks sliding into the opposite corner, where results are measured carefully but nobody asks who bears the cost. Both corners produce bad decisions, for different reasons.

Historically, Swedish public administration has had an explicit ambition to sit in the principled corner: evaluation, follow-up and the principle of public access are precisely tools for testing whether well-intentioned measures work. The question is not whether the ambition exists, but whether the tools are actually used — and that is an empirical question with Swedish answers.

How an open society should deal with those who want to abolish openness is not a 2020s invention. Karl Loewenstein formulated the concept of ”militant democracy” in the American Political Science Review in 1937, against the background of democratic states in Europe having fallen through their own procedures.

Karl Popper described the paradox of tolerance in 1945: if a society is unlimitedly tolerant even of the intolerant, tolerance will eventually be destroyed. Popper was equally careful that the paradox does not justify banning views one dislikes — countermeasures are warranted only once the other side has rejected rational argument.

Article 17 of the European Convention on Human Rights codified the principle in 1950: the convention's rights may not be invoked to destroy the rights of others. It is a legal limit on tolerance, written by people who had just seen what the absence of such a limit can cost.

This tradition gives the question legitimacy. It does not give Saad's thesis support. Loewenstein and Popper speak of institutional and legal limits against anti-democratic movements — not of citizens' emotional lives as a cause of political outcomes. Invoking them in support of a critique of empathy is changing the subject.

Here the analysis moves from ideas to Swedish sources. The question is not whether ”suicidal empathy” exists in Sweden — the concept is not measurable — but whether there are documented cases where Swedish institutions failed to test the consequences of their measures. There are, and they are named and public.

In 2025 the Swedish Schools Inspectorate, on the government's instruction, reviewed teaching on sexuality, consent and relationships in 60 schools. It concluded that 23 of 30 compulsory schools and 26 of 30 upper-secondary schools need to raise quality, that pupils are given limited opportunity to develop knowledge about honour-related violence and oppression, and that teachers report uncertainty about teaching sensitive questions. The last point is of interest to the debate — but the agency attributes it to uncertainty and a lack of a coherent approach, not to concern for perpetrators.

In 2023 the Swedish National Audit Office reviewed state grants to civil society and found that the system is not designed so that it can effectively prevent funds going to organisations that fail to meet the requirements. This is a documented control failure. The report describes it as a governance and rule-design problem — not as a result of caseworkers' feelings.

Parliament subsequently legislated a unified and clearer democracy condition for state grants to civil society, in force from 1 January 2025. That is an example of the system correcting itself, which is hard to square with a picture of institutions incapable of self-examination.

The 2024 ESO report Eloquent silence? (Widmalm et al.) shows that 53% of respondents in 2021–2022 agreed with the statement: "In today's political climate I cannot openly express my views because others might find them offensive." The share varies sharply by party sympathy — highest among Sweden Democrat sympathisers (83%). This is the same figure and the same survey question we report on our page on freedom of expression, and it is the Swedish finding closest to Saad's mechanism.

But it is self-reported perception in a survey, not measured behaviour in decisions — and the gap between them is the entire causal question.

In sum: the Swedish shortcomings are real, documented and serious enough to warrant action. None of the sources uses empathy as an explanation. Laying that explanation on top is an interpretation, and it should be presented as one.

The clearest documented case where fear of appearing prejudiced coincided with inaction is not in Sweden but in the United Kingdom. Alexis Jay's independent inquiry in 2014 established that at least 1,400 children were sexually exploited in Rotherham between 1997 and 2013, and that the authorities had repeated knowledge of it.

The Jay report is unusual in that it addresses the question of motive explicitly. It describes several staff saying they were uncomfortable raising the ethnicity of perpetrators, and that some were advised not to. That is documented silence linked to fear of being seen as racist — and it cost children their childhoods.

Louise Casey's 2015 inspection deepened the picture and described a culture of denial within the council. Together the two reports constitute the strongest single evidence that the mechanism Saad describes can occur in reality.

But the limits must be stated just as clearly. The Jay report also describes systematic failures with nothing to do with ethnicity or empathy: poor leadership, the dismissal of victims as ”troubled youths”, weak cooperation between police and social services, and a pervasive failure to believe the girls. Extracting one of several documented reasons and making it the whole explanation misrepresents the report.

And above all: a British case shows the mechanism can occur. It does not show it is widespread, and it shows nothing at all about Swedish conditions. An existence proof is not a frequency estimate.

A serious examination must also turn on the criticism. The concept of ”suicidal empathy” carries three built-in risks, and they are not hypothetical.

The first is that criticism slides from the empathy to the group. An argument that formally concerns untested benevolence is in practice almost always deployed about concern directed at a specific group — most often migrants or minorities. The test is simple: if the same reasoning would not be accepted when applied to another group, empathy is not the actual subject.

The second is that the concept is immune to disproof. Because it has no operationalisation, any outcome can be read as confirmation. A society doing well has not yet been struck; one doing badly has been. That is rhetorically powerful and analytically worthless.

The third is that the criticism can become its own mirror image. If ”empathy” becomes the thing under suspicion, a new moral coding emerges in which hardness signals judgement and compassion signals naivety. One has then swapped one moral filter for another without any testing having improved. That is the bottom-right corner of the model above.

There is also a simple empirical objection: most of the institutional failures Saad's adherents point to have more parsimonious explanations. Vague rules cannot be followed up. Overloaded agencies lack capacity. Conflicting objectives that no one has ranked get resolved case by case by individual officials. None of these explanations requires assuming anything about anyone's feelings — and they have the further merit of being fixable.

If the concept has any value, it is as a diagnostic lens on individual decisions, not as a diagnosis of a society. The questions below are framed so they can be asked of a concrete case — a measure, a rule, an agency decision — and so they work regardless of the asker's political views.

Note that question five is the most important. If an outcome can be explained by conflicting objectives, resource constraints or unclear rules, nothing needs to be assumed about feelings at all — and the simpler explanation should be tested first.

The distortions of empathy are real: it is partial, it does not scale with the extent of suffering, and it can amplify polarisation rather than dampen it. That is well documented and should be taken seriously by anyone designing policy.

Swedish institutions have documented shortcomings in control and follow-up in several areas, established by the National Audit Office, the Schools Inspectorate and ESO. That too should be taken seriously, and several of the shortcomings have already led to tightened rules.

But the bridge between these two observations does not exist in data. No Swedish study measures whether empathic norms affect institutional outcomes, and the claim can therefore be neither confirmed nor refuted. That is a gap in knowledge, not proof in either direction.

”Suicidal empathy” therefore works best as a question and worst as an answer. The question — do we test the consequences of what we do out of benevolence? — is legitimate, old and necessary. The answer, that compassion causes national decline, is a metaphor that cannot bear the weight placed on it.

Sverigefakta's position is accordingly: use the question, reject the conclusion, and demand the same standard of evidence from the critique of empathy as from what it criticises.