Why Mental Health Apps May Harm the People They Aim to Help

The App That Was Supposed to Save You Might Be Part of the Problem

In 2020, as the world locked down and anxiety spiked, a quiet promise circulated through tech blogs and mental health conferences: download this app, and you will get the care you need. No waiting lists. No insurance forms. No stigma. Just a smartphone, a few taps, and a guided path toward feeling better.

It was a beautiful story. And like many beautiful stories about technology, it turned out to be dangerously incomplete.

Sachin R. Pendse, Daniel Nkemelu, Nicola J. Bidwell, and Sushrut Jadhav, researchers at Georgia Tech and University College London, spent two years examining the hidden architecture of digital mental health tools. What they found is uncomfortable: the very design of these apps, the way they categorize human suffering, the data they collect, and the assumptions baked into their code may be causing real harm to the people they claim to help (Pendse et al., 2022).

Not because the apps are buggy. Not because they are poorly marketed. But because they are built on a foundation that was never designed for most of the people using them.

What Gets Lost When You Turn Pain Into Data Points

The problem starts with a simple question that every mental health app must answer: how do you measure suffering?

Most apps use standardized screening tools. You answer questions about your mood, your sleep, your appetite. The algorithm assigns you a score. The app suggests a breathing exercise or a cognitive behavioral therapy module. Done. Next user.

But here is what Pendse and his colleagues discovered: this process does not just measure distress. It actively reshapes it. When you force a person's lived experience into a rigid classification system, you erase everything that does not fit (Pendse et al., 2022).

Think about what that means for someone whose depression is tied to poverty, racism, or historical trauma. The app asks: "How many days in the past two weeks have you felt hopeless?" It does not ask: "What happened to your family during the war?" It does not ask: "Do you feel safe walking down the street?" It does not ask: "Is your landlord threatening to evict you?"

These are not edge cases. They are the majority of the world's population. The authors draw on decolonial thought to argue that Western mental health frameworks, which dominate app design, were developed in specific cultural contexts for specific populations. Exporting them globally through technology is not neutral. It is a form of epistemic violence: telling people that their way of understanding their own suffering is wrong (Pendse et al., 2022).

The methodology here is worth understanding. Pendse and his team did not run a controlled trial. They conducted a critical analysis, examining the underlying power relations embedded in digital mental health technologies. They analyzed how identity based algorithmic bias operates in these systems. They interviewed experts and reviewed the literature on decolonial theory. This is not a study that tells you "app X caused harm in 30% of users." It is a study that asks a deeper question: what assumptions are we making about human beings when we design these tools, and whose interests do those assumptions serve?

The Algorithm Does Not Know Your Context

Here is where it gets concrete.

Imagine two people open the same mental health app. Both are experiencing suicidal thoughts. One is a white college student in a wealthy suburb whose family has health insurance and access to therapy. The other is an Indigenous teenager on a reservation where the nearest mental health clinic is 200 miles away and the historical trauma of forced assimilation is still alive in community memory.

The app treats them identically. It screens for the same symptoms. It offers the same coping strategies. It categorizes them into the same diagnostic boxes.

But their suffering is not identical. It cannot be, because suffering is never just a neurological event. It is always shaped by history, by power, by place. Pendse and his colleagues argue that digital mental health tools, by design, strip away this context. They treat the individual as a standalone unit of analysis, disconnected from the structural forces that produce and maintain distress (Pendse et al., 2022).

This is not an accident. It is a design choice. The authors found that most digital mental health platforms are built on what they call "the treatment paradigm": a model focused on symptom reduction, efficiency, and scalability. This model works well for mild to moderate anxiety and depression in populations that match the clinical trials. It fails catastrophically for anyone whose pain is rooted in systemic oppression, because the app cannot fix the system. It can only offer you a breathing exercise and hope you do not notice the difference.

The Data You Give Away Might Be Used Against You

There is another layer to this that is deeply unsettling.

Mental health apps collect vast amounts of sensitive data: your mood patterns, your sleep cycles, your location, your social media activity, sometimes even your voice recordings or typing speed. The apps promise this data is private. But Pendse and his team point out that these promises are built on a Western model of privacy that does not account for how data can be weaponized against marginalized communities (Pendse et al., 2022).

Consider what happens when your mental health data is shared with an employer, an insurance company, or a law enforcement agency. In many countries, there are few legal protections against this. The authors cite examples of digital mental health platforms being used to deny insurance coverage, to justify workplace discrimination, and to monitor immigrants and refugees.

The problem is not just that data leaks. The problem is that the data itself is collected through a system that was never designed to protect the people who are most vulnerable. The authors argue that digital mental health tools create power differentials: the app knows everything about you, and you know almost nothing about how the app works or where your data goes. This asymmetry is not a bug. It is a feature of the business model (Pendse et al., 2022).

The Colonial Roots of the "Self Care" Revolution

This is the part that might make you uncomfortable.

The authors trace the genealogy of digital mental health back to colonial psychiatry. They show how Western mental health frameworks were historically used to pathologize resistance to colonial rule. People who fought against oppression were diagnosed as mentally ill. Their suffering was individualized, stripped of its political meaning, and treated as a medical problem (Pendse et al., 2022).

Today's mental health apps do the same thing, just with better design and more funding. When an app tells you to "practice self care" instead of organizing against an unjust system, when it frames your anger at oppression as a symptom to be managed, when it offers you a meditation track instead of solidarity, it is reproducing a colonial logic. The authors call this "therapeutic governance": using mental health discourse to pacify dissent and maintain existing power structures (Pendse et al., 2022).

This is not to say that meditation or cognitive behavioral therapy are inherently bad. They can be genuinely helpful. But the authors argue that when these tools are delivered through digital platforms that are owned by corporations, funded by venture capital, and designed to maximize user engagement, they become something different. They become mechanisms for extracting value from human suffering while doing nothing to address its root causes.

What Would a Decolonial Digital Mental Health Look Like?

Pendse and his colleagues do not stop at critique. They offer a vision.

A decolonial digital mental health, they argue, would center lived experience over rigid classification. It would be conscious of structural factors that influence mental wellbeing. It would be fundamentally designed to prevent power differentials that stop people from having agency over their care (Pendse et al., 2022).

What does this look like in practice?

The authors suggest several concrete design principles:

• ▸Community ownership: Apps should be governed by the communities they serve, not by corporate shareholders. This means transparent algorithms, open source code, and decision making power held by users.

• ▸Contextual assessment: Instead of standardized screening tools, digital mental health tools should use narrative based approaches that allow people to tell their story in their own words, on their own terms.

• ▸Structural awareness: Apps should help users understand the social and political determinants of their distress, not just their individual symptoms. This might mean connecting users to mutual aid networks, legal support, or housing assistance.

• ▸Data sovereignty: Users should have complete control over their data, including the right to delete it permanently. Data should not be sold, shared, or used for research without explicit, informed, revocable consent.

• ▸Pluralism: There is no single way to understand mental distress. Digital tools should support multiple frameworks, including Indigenous healing practices, community based care, and peer support networks.

The authors are careful to note that these are not technical fixes. They are political choices. And they require a fundamental shift in how we think about mental health technology: from a tool that delivers treatment to a platform that supports healing.

What the Research Does Not Prove

It is important to be honest about the limits of this analysis.

Pendse and his colleagues did not conduct a randomized controlled trial. They did not measure the effectiveness of any specific app. They did not prove that digital mental health tools always cause harm. In fact, they acknowledge that many people find genuine relief through these platforms.

What they did was identify a pattern: the underlying design logic of most digital mental health tools is structurally aligned with the interests of power, not with the needs of the most vulnerable users. This is a claim about architecture, not about individual user experience. It is possible for a person to benefit from an app that is, in its broader design, harmful to others.

The open question is whether these two things can be separated. Can we design digital mental health tools that are both effective and decolonial? Or is there something inherent in the technology itself that reproduces the problems the authors identify? The paper does not answer this definitively. It issues a warning and a challenge.

What This Actually Means

• ▸The next time you recommend a mental health app to someone, ask yourself: does this tool acknowledge the structural context of their suffering, or does it treat them as an isolated individual whose problems can be solved with a breathing exercise?

• ▸If you design or fund mental health technology, the single most important question is not "does it reduce symptoms?" but "who holds power in this system?" If the answer is a corporation or a venture capital firm, the design is already compromised.

• ▸For users: demand transparency. Ask what data is being collected, who owns it, and how it might be used against you. If the app cannot answer these questions clearly, do not use it.

• ▸For researchers: stop treating "efficacy" as the only metric that matters. Measure power dynamics. Measure community ownership. Measure whether the tool increases or decreases user agency over their own care.

• ▸The most radical thing a mental health app could do is not to help you manage your symptoms. It is to help you understand that your suffering is not your fault, that it is connected to systems larger than yourself, and that healing might require collective action, not individual optimization.