The free speech problem is (still) a crisis of confidence

There is a version of the campus free speech story that the political movement behind the Higher Education (Freedom of Speech) Act 2023 has told for a decade.

Jim Dickinson

Jim is an Associate Editor (SUs) at Wonkhe

What the research actually found

The headline OfS leads with is reassuring. 84 per cent of students say they are aware they have some rights to freedom of speech, and 76 per cent think it is well protected on their campus.

Dig one layer down and the reassurance thins. Only 16 per cent say they know “a lot” about those rights, 49 per cent know “a little”, another 20 per cent are aware they have rights but know nothing about them, and 11 per cent aren’t aware at all – so close to a third have no substantive grasp of the rights the headline says they hold.

Only 52 per cent recall their institution communicating anything to them about those rights at all. “High awareness” turns out to mean a vague sense that protections exist, not knowledge of what they are.

A key finding is about setting. Students feel markedly freer to discuss challenging or controversial topics in small, in-person, unrecorded spaces than in large or online ones, and the gradient is steep: 71 per cent feel free in in-person seminars, falling to 55 per cent in in-person lectures and 46 per cent in online lectures.

The dominant barrier almost everywhere is fear of social consequences – peers thinking poorly of them, social ostracism, unwanted attention – rather than fear of institutional punishment. This is not a story about a censorious regulator-shaped problem awaiting a regulator-shaped solution. It is really a story about climate.

One caveat – there is as ever no general-population comparison anywhere in the report. We cannot say whether students self-censor more, less or about the same as the rest of the country, which means every percentage here floats free of the only benchmark that would tell us whether campus is unusually chilled or merely as chilled as everywhere else.

The seminar v the lecture hall v the cloud

We pretty much made this argument back in early 2023, when the prevailing national narrative was all cancel culture and no-platforming. The contrarian read at the time was that the bulk of free speech negativity wasn’t ideological at all but a mix of subject, teaching format and, above all, confidence and anxiety.

The OfS data, two and a half years later, has effectively caught up with that argument – its seminar-versus-lecture thing is exactly the format-and-massification effect we described, only now with weighted percentages attached.

What OfS adds is a mechanism our 2023 work only gestured at – recording and permanence. Online and recorded settings feel riskier specifically because what you say persists, can be screenshotted, taken out of context and spread. The report draws the inference explicitly, suggesting that worry about being “on the record” may also explain students’ marked reluctance to commit controversial views to a university or student publication (just 42 per cent feel free to do so, against 67 per cent in student groups and societies).

This cuts awkwardly across another good-practice agenda entirely – lecture capture and online delivery, pushed hard on accessibility and cost grounds, are functioning here as speech suppressants. Nobody has been treating recording as a free speech problem. The data says that for students it is one.

This is where our forthcoming Friends have benefits research becomes relevant, and where the OfS material both corroborates and sharpens. That work reads freedom of expression as fundamentally a question of classroom climate – whether your classmates would judge you – rather than of policy, and the OfS data is clean validation.

But the recording mechanism is where a belonging reading is incomplete. Part of why the seminar feels free is the simple absence of a paper trail, which has nothing to do with the strength of your cohort and everything to do with whether anyone is pressing record.

There is a subject pattern too – liberal arts, humanities and law students feel freest, while STEM students more often report they have fewer opportunities for contested discussion because their subjects have right-or-wrong answers, not because they are being silenced.

It’s worth flagging the measurement risk buried in that. If the regulation is ever monitored through climate surveys like this one, STEM-heavy and specialist providers will score low on “free speech” largely because their disciplines throw up fewer contested questions – an artefact of subject mix, not suppression. We have seen enough B3 and TEF benchmarking rows to know where that road leads.

Two gender gaps, not one

OfS notes, as these reports often do, that men feel freer than women – more likely to rate protection “very” good (26 against 19 per cent) and freer on social media (58 against 50 per cent). What the data does usefully, and what the insight brief rather skates over, is decompose why.

Women who self-censor overwhelmingly cite a lack of confidence or not feeling knowledgeable enough – in in-person lectures, 50 per cent of women against 35 per cent of men. Men who self-censor more often fear backlash and social ostracism (68 against 57 per cent on ostracism specifically). These are not the same phenomenon wearing one label.

Women’s pattern is, at base, a confidence-and-belonging problem dressed in free speech clothing – and a condition of registration, a code of practice or a rebuttable presumption barely reaches it. A meaningful share of what gets measured and reported as “self-censorship” is not censorship in any rights-based sense – it is diffidence, and the remedy is pedagogy and participation.

Our 2023 argument said “confidence.” Our forthcoming Friends have benefits work will sharpen that into peer climate and fear of judgement. The OfS evidence lands squarely on the relational end – fear of social consequences, not generalised individual anxiety – which suggests our original “crisis of confidence” framing was under-weighting how relational the fear really is. The regulator’s own data matures the diagnosis in precisely the direction our belonging work took it.

Who self-censors

The popular framing runs roughly – a progressive student majority chills gender-critical or right-leaning dissent. The OfS data makes a mess of that. Self-censorship here is plainly bidirectional – a Conservative party member who admits pretending to be “more left wing” in front of peers, pro-Palestinian students who fear being labelled pro-terrorism, students with pro-Jewish views who fear being branded “pro-genocide”, religious students who fear ridicule.

The unifying feature is a dominant local view that can’t safely be challenged, not a particular politics. That chimes with the conformity strand in our forthcoming research, where perceived left-wing peer dominance and institutional suppression of pro-Palestine activism appear as two faces of the same dynamic, frequently in the same testimony.

Here is the inversion OfS doesn’t draw, and arguably can’t, given the politics of the agenda it serves. The students who self-censor most – ethnic minority students, international students, religious minorities – are also the students most likely to want their university to prioritise freedom of speech. Asked whether their institution would put free speech above students and staff feeling safe, white students more often expect safety to win (62 against 53 per cent), while ethnic minority students are more likely to want free speech prioritised (22 against 16 per cent), as are international students.

The constituency the free speech debate is usually framed as protecting students against turns out to be the constituency asking for more of it. The finding ought to detonate the lazy version of the argument, and so it won’t be a surprise that it is nowhere in the brief’s narrative.

It gets more structural still. Among ethnic minority students who self-censor in lectures, 26 per cent fear legal consequences, against 14 per cent of white students, alongside 56 against 38 per cent fearing academic consequences. “Legal” is doing heavy work there – immigration enforcement, Prevent, harassment law reaching into the seminar room.

And the international-student material is the darkest strand in the report – students describing UK campuses as “safer spaces” than home in one breath and, in the next, fear of visa repercussions, suspension and deportation over (especially pro-Palestinian) speech, leading some to conceal their identity at protests. What the report can’t do anything about is that the principal chilling mechanism for its most-affected group sits with the Home Office, not the provider.

The coursework blind spot

Fear of social consequences dominates in almost every setting, with one exception. In coursework, the fear that dominates is academic consequences, cited by around three-quarters (74 per cent) of those who feel uncomfortable. The entire public debate fixates on protests, visiting speakers and classroom rows. The sharpest chill the data finds is in the graded work that determines the degree classification – students trimming what they argue in assessment because they fear it will cost them marks.

That is a question about academic freedom of the student, about marking culture and assessment design, and it is almost entirely untouched by a frame built around speech events and institutional codes. It is also a far more tractable thing to raise with an institution than another row about a speaker.

What students say would help

Asked what would encourage open discussion, students reach first for staff stating the class is a “safe space” (43 per cent), clear ground rules set at the outset (42 per cent), safeguarding against online harassment (38 per cent) and more student-led discussion spaces (35 per cent).

“Safe space” is the precise phrase the political movement that produced this Act has spent a decade deriding as the enemy of open debate – and it is the single most-supported thing students ask for. Notably, too, more students want student-led spaces than university-led debating events (35 against 30 per cent), and the intervention they want least, by a distance, is “staff not expressing their personal views” (11 per cent). That last number guts the academic-neutrality strand of the project. Students do not want silenced, viewpoint-neutral academics – they want skilled facilitators who set boundaries and intervene when discussion tips into harassment. They want better teaching.

Here the brief’s framing parts company with its own evidence. The OfS leans on the finding that 86 per cent of students think “at least one” intervention could help, presenting it as appetite for reform. But that aggregates support across nine options, several of them weak, with no single measure commanding a majority, and it tidies out of view the 4 per cent who think nothing would help and the 11 per cent who don’t know.

Bundled, it reads as a mandate. Disaggregated, it is tepid. The brief uses the bundle. More consequentially, the conclusion lands hard on a causal claim – that ambiguity about policy boundaries itself causes self-censorship, more than the rules themselves do. To be fair, that is plausible, and a sharper version of the case for clear ground rules. But it is drawn from participant suggestion, not a tested relationship – and it is, conveniently, exactly the proposition the condition of registration needs to be true. When a regulator’s flagship finding is the one that most neatly justifies the instrument it is about to deploy, a sceptical reading is the responsible one.

There is a deeper one-sidedness in the instrument itself. The survey asks how free students feel to speak. It does not seriously measure the other half of the ledger – their experience of being on the receiving end, of harassment, intimidation, being shouted down. Yet the condition being built on top of it, like the existing E6 condition on harassment and sexual misconduct, exists precisely to balance freedom to speak against freedom from harm.

An evidence base that measures only one side of a balance is a thin foundation for regulation whose entire job is to hold the two in tension. And underneath all of it sits the strongest unstated assumption in the enterprise – that a provider-directed condition is the right instrument for a phenomenon the regulator’s own evidence shows is driven chiefly by peer dynamics, the wider political and media climate, assessment design and immigration enforcement – none of which providers fully control.

The clarity students want

The worldview running through the data is a balancing one – draw a clear line between lawful free speech and harmful behaviour, protect me from offence, hate and harassment.

That instinct is not, in itself, at odds with where the law has landed. In R (University of Sussex) v OfS, Justice Lieven rejected the regulator’s absolutist position that any restriction of lawful speech amounts to a breach, and confirmed that providers can restrict lawful speech where doing so is proportionate.

The catch, as we covered when the judgment landed, is that proportionality alone is no longer enough – a provider must also show that allowing the speech to proceed unimpeded was not reasonably practicable, a higher and considerably messier test. So the students reaching for clear, predictable rules that shield them from lawful-but-offensive speech are reaching for certainty in exactly the area the law has just made more uncertain and more litigable. The bright line they most want is the thing the post-Sussex landscape is least able to give them.

So as is becoming common with these “polling plus insights” combos from OfS, the most useful thing in this report is not the conclusion the regulator drew from it. It is the evidence the regulator leaves lying around – that the chill is real, that it is mostly social rather than institutional, that it falls hardest on the students the debate usually ignores, and that the cheapest things that would help are within the gift of any course team that cares to set clear expectations and build the confidence to use them.

The questions worth keeping a list of are evidential, not procedural – whether students self-censor any more than the general public, which the report gives us no way to know; what actually happened to the students referenced as disciplined or suspended over Palestine-related speech, which the report banks as a chilling effect and never examines; and whether any of the favoured interventions actually work, given that the survey measures only what students think might help, not what does.

You don’t need a new condition of registration to start finding out.

Wonk Corner | 28/05/26