The Crisis Narrative Problem in Mental Health (And Why Builders Should Audit It)

Crisis language works. It mobilizes funding, attention, and urgency in a field that has historically struggled to get any of the three. It's also, according to Dr. Ben Locke, Chief Clinical Officer at Togetherall and founder of Penn State's Center for Collegiate Mental Health, quietly become part of the system it claims to describe, in ways that should make anyone building mental health products uncomfortable.

Locke's argument, delivered to a room of mental health builders and clinicians, was not that distress is fake or that care is unnecessary. It was narrower and more useful than that: that the language, measurement, and marketing built around "crisis" can shape identity, survey responses, institutional priorities, and product strategy, often invisibly, and that most organizations have never audited where their own work depends on it.

The crisis narrative as a design constraint

The macro story that has built up over roughly two decades runs like this: mental health is worsening, services are overwhelmed, and the solution is to identify more people in distress and route them into care. That story is not baseless. It has also, in Locke's framing, narrowed what builders notice and reward, in a specific and testable way.

The risk is pathologizing ordinary human experience. If every difficult feeling gets translated into symptom language, people can start to absorb those labels as identity rather than as a passing state. That reframes the product question from "are we reducing suffering?" to something closer to "are we making more people fluent in diagnostic self-description?" Those are not the same outcome, and a product optimizing for the second can look successful (more engagement, more assessments completed, more people "identified") while doing very little for the first.

Language spreads, and it doesn't stop at the extremes

A central part of Locke's argument concerns contagion: the well-documented phenomenon where humans learn emotional cues, interpretive frames, and social scripts from each other. He pointed to tic outbreaks and social-media-mediated symptom spread as the vivid, extreme examples most people already accept, then asked the harder question: why would we assume contagion only operates at the extreme edges of mental health, rather than as a general feature of how people learn to interpret and describe their own internal states?

The practical implication for anyone building a product is direct. Copy, onboarding questions, check-in prompts, surveys, and marketing claims are not neutral wrappers around an intervention. They can change how a user interprets themselves before any clinical or support content has even begun. A check-in that frames a normal bad day in clinical language is not a passive measurement tool; it's actively shaping the user's self-concept, for better or worse.

What happens to your metrics when the vocabulary changes

Locke used the PHQ-9 to make this concrete during audience questions, and it's worth sitting with. The instrument was designed before psychological vocabulary became as mainstream and widely understood as it is today. As that vocabulary spreads through a population, endorsement rates on a self-report instrument can shift, not because underlying distress changed, but because interpretation, willingness to report, and ordinary conversational idiom have all changed alongside it.

This does not mean self-report data should be discarded. It means trend lines need real humility. A rising PHQ-9 average across a population, or across your own user base, could reflect worsening symptoms, growing willingness to endorse items that were previously under-reported due to stigma, or simply a population that has become more fluent in describing feelings using clinical terms. Those three explanations point to entirely different product and business decisions, and most organizations reporting a "rise in anxiety" or "rise in depression" in their user base have not actually distinguished between them.

The audit, in practice

Locke's operating advice was specific: go back through the artifacts your organization already produces (pitch decks, product goals, marketing copy, intake surveys, referral logic, outcome claims) and check where each one depends on crisis framing to make its case. For each dependency found, ask two questions. What evidence actually supports this specific claim, for this specific population? And what happens, to users, to funders, to your own decision-making, if the claim turns out to be overstated?

This is not an argument for abandoning urgency where urgency is genuinely warranted. Crisis intervention, suicide risk screening, and acute safety infrastructure are not the target of this critique, and treating them as optional would be a serious misreading. The target is the much larger volume of ordinary product and business language that borrows crisis urgency without the underlying evidence to support it: growth narratives that lean on "the mental health crisis" without defining which population, which condition, or which specific gap they're addressing; engagement metrics framed as "more people getting help" without distinguishing help-seeking from actual improvement; survey results presented as trend evidence without accounting for the vocabulary shift Locke describes.

Key takeaways from the session

The crisis narrative can unlock attention and funding, but it can also shape identity, demand, measurement, and product incentives, often at the same time and through the same mechanism. Mental health language functions like an intervention: labels, copy, surveys, and screening flows can change how people understand themselves, independent of any clinical content that follows. Awareness and identification are not the same outcome as improved outcomes, and conflating them is one of the most common ways crisis-framed products look successful without actually helping anyone. Survey trend lines need careful interpretation whenever public familiarity with psychological language is changing, which, in 2026, it is, rapidly. The safer posture is not denial that distress exists. It is humility about causality, about what a given metric is actually measuring, and about the unintended consequences of language that has become load-bearing for a business model rather than descriptive of a clinical reality.

As Locke put it directly: "If we begin to take every aspect of the human experience, and label those normative experiences as problems, then we have pathologized the normative human experience." His closing challenge to the room, worth sitting with regardless of what you're building, was simpler still: "Where else are you relying on the crisis narrative?"

This is a useful companion question to why AI is expanding, not shrinking, the total demand for mental health care: expanding help-seeking is good, but only if the measurement underneath it can actually distinguish more people getting genuinely better from more people simply becoming fluent in describing themselves as unwell.

---

Citt.ai's clinical assessment tools are built with this measurement humility in mind: tracking real symptom trajectories over time, not just point-in-time scores that can be confounded by shifting vocabulary. Explore assessment and progress tracking.