Building Trust in AI for Healthcare: Overcoming Emotional Assurance Barriers

AI is becoming the first layer of engagement in healthcare, powering everything from symptom-checkers to therapy bots. However, adoption is outpacing trust. In Singapore, a highly digitized and regulated environment, AI is everywhere, with 80% of residents using it. Yet, trust drops when the advice involves sensitive areas like mental health. This trend is common across South-East Asia, where one in four people in Indonesia and Hong Kong has tried an AI mental health tool. However, concerns about safety and empathy still persist. In the US, nearly 60% of Americans feel uneasy about AI‑aided diagnoses. The issue is not technical accuracy; it’s a crisis of emotional assurance.

This trust gap stems from three systemic issues:

1. Opacity: Patients and clinicians often cannot see how AI risk scores are generated. This lack of transparency creates a trust gap. It’s not just an inconvenience; it can lead to patient harm. The OECD AI Incidents Monitor reports cases where flawed AI design led to biased outcomes, like one system prioritizing white patients over black patients. Regulatory audits have confirmed that over 90% of FDA-approved AI devices don’t disclose essential information about their training data or architecture. When users can’t see how decisions are made, safety becomes guesswork.
2. Accountability: The ITU’s AI Governance Report explains how responsibility is divided among developers, hospitals, and ministries when private AI models enter public health workflows. This fragmentation creates an accountability gap. When an error occurs, no one takes ownership. Patient grievances often get lost in bureaucracy with no one obligated to explain, investigate, or fix issues, eroding trust.
3. Human control: When AI suggests a diagnosis first, it shifts the traditional clinical workflow. Research shows that when clinicians engage with AI-proposed diagnoses, they must verify them. Their acceptance depends on the model’s ability to explain its reasoning. However, assuming human oversight means little without clear, operational checkpoints. Clinicians must read and either accept or reject AI diagnoses.

Why Perceived Safety is the Next KPI

In healthcare, digital trust is essential for effectiveness. Studies show that unease with AI reduces engagement and increases dropout rates, even when accuracy is high. Users disengage not because the model is wrong, but because the experience feels unsafe.

AI cannot feel empathy, so trust cannot rest on its human-like connection. Instead, trust in healthcare AI builds on respect for the patient’s vulnerability, a dynamic defined by predictability, clarity, and user control. To build this trust, three elements must be clear:

1. Which data is used: A 2024 study in Nature Medicine shows that unclear data flows reduce trust in medical advice. When patients don’t understand how data is used, they are less willing to share sensitive information.
2. How decisions are made: A recent audit by Nature found that over 90% of FDA-approved medical AI devices fail to disclose basic information about their training data or architecture. This lack of transparency weakens trust in the accuracy of AI recommendations.
3. When humans are involved: Studies on AI responsibility gaps show that these gaps often result from diffused accountability. Patients need clear pathways to request an explanation, review, or correction of an AI decision. This process helps restore accountability and strengthens trust.

Trust by Demonstration in South-East Asia

Health systems are shifting from designing trust on paper to showing it in practice. These reforms make safety observable, shifting trust from a promise to proof:

• Singapore has introduced reforms that require audit trails and cybersecurity measures before deploying AI diagnostics. Cross-border collaboration has also increased, with the Singapore-Malaysia Medical Device Regulatory Reliance Programme speeding up evaluations through shared oversight.
• Indonesia is integrating AI assurance principles into frontline care with its BPJS Digital Health Transformation Strategy, which creates the necessary infrastructure for future AI-supported triage.
• Malaysia has prioritized regional cooperation on ethical AI during its ASEAN chairmanship. This collaboration ensures safety and traceability are core to the user experience.
• Hong Kong is developing the foundational infrastructure for trusted data-sharing. A recent initiative led by the Chinese University of Hong Kong and Hong Kong Science Park is building the region’s first cross-border medical data space, ensuring secure and credible data handling.

Introducing Minimum Viable Assurance

Building trust at scale requires minimum viable assurance. Three metrics can help achieve this:

1. Clinician override rates: Evidence from real-world deployments shows that tracking how often clinicians reject AI recommendations provides a practical signal of model reliability. A study found that override rates were just 1.7% for trustworthy, transparent AI predictions compared to over 73% for opaque models. This shows that override rates are a strong safety indicator.
2. Audit trail visibility: The WHO on AI for Health mandates mechanisms for audit and human oversight. This is foundational for model-level logging and verifiable accountability, making it something users can verify.
3. Patient comprehension scores: Clarity directly affects whether patients follow recommendations. “Teach-back” checkpoints, where patients confirm their understanding, offer a tangible way to ensure safety and trust.

A New Policy Agenda for Leaders

As AI becomes more common in health systems, a new priority must guide governance: patient trust as a core performance indicator. This means shifting from evaluating only technical accuracy to focusing on human confidence.

Today, assessments focus on model accuracy and efficiency. However, real-world adoption depends on whether tools feel safe and fair to use. To close this gap, policy must mandate continuous trust assurance alongside technical validation, moving beyond one-time audits to ongoing monitoring of real-world impact.

Key actions include:

• Operationalizing equity: Track and address lower uptake among older and lower-income populations, even for free tools.
• Building visible recourse: Establish clear pathways for patients to challenge AI decisions, turning regulatory compliance into tangible user control.
• Addressing digital discomfort: In mental health, where provider and patient hesitancy can turn AI from a bridge into a barrier, widening access gaps.

The future of digital health will be shaped by confidence in clinics, not just algorithmic performance in labs. To make AI truly effective, trust must be engineered into every interaction.

READ: Kenya Signs Historic $2.5 Billion Health Agreement with the US