Why Our Trust in AI Systems Is Eroding and How to Rebuild It

As artificial intelligence continues to evolve, so does our relationship with it. I’ve seen firsthand how AI’s potential can inspire hope and skepticism in equal measure. While it holds immense promise, a series of high-profile failures and oversimplified solutions have steadily eroded public trust in its capabilities. I’ll walk through why trust in AI systems is diminishing, and more importantly, how I believe we can rebuild it through responsible, transdisciplinary collaboration.

Trust in AI systems is eroding due to public failures, overhyped technological solutions, and a blind faith in scaling models without addressing inherent errors. From my perspective, restoring that trust means emphasizing transparency, accountability, and a deeper, more integrated strategy, one that draws from a broad range of scientific and human-centered disciplines.

KEY TAKEAWAYS:

• Public failures matter. High-profile breakdowns have accelerated the erosion of trust in AI systems, especially in critical sectors.
• Scale has limits. Massive models still carry error; scaling alone will not solve the problem of trust.
• Engineering is necessary but not sufficient. Scientific rigor and ethical design must complement technical development.
• Transparency and accountability are non-negotiable. Explainable AI and strong governance frameworks are essential for public confidence.
• Transdisciplinary collaboration is the way forward. Building trust requires integrated thinking across fields, communities, and agencies.

In the sections that follow, I’ll explore the main factors driving the erosion of trust in AI, highlight why current approaches often fall short, and outline how transdisciplinary collaboration can offer a more effective path forward.

Why AI Failures Have a Larger Impact Today

AI failures aren’t new, but their impact today is broader and harder to ignore. I still remember when Microsoft’s Tay chatbot went off the rails within hours, the public reaction was swift and lasting. Similar failures in healthcare, hiring, and autonomous systems have shown how flaws in AI can create serious real-world consequences. What’s changed isn’t just the technology, but how deeply it’s embedded in our lives. Half of U.S. adults now say they’re more concerned than excited about AI, up from 37 % in 2021, a clear sign of growing public anxiety. As AI takes on more critical decision-making roles, even small issues, bias, opacity, unpredictable outputs can erode trust in AI systems.

The Myth of Bigger AI Models: Why More Data Isn’t Always the Answer

There’s a widespread belief that with enough compute, data, and engineering, AI will simply get better. But large models, no matter how advanced, are still statistical, and statistical models carry error. I’ve seen scale amplify uncertainty rather than reduce it. Massive datasets don’t eliminate bias or unpredictability; they often obscure it. And when a system’s behavior becomes too complex to explain, trust in AI systems begins to erode. Without grounding in decision science or economics, organizations risk overlooking critical tradeoffs. In my experience, there’s a point where bigger stops being better. The focus needs to shift toward designing systems that are transparent, verifiable, and grounded in scientific understanding.

Engineering Alone Won’t Solve AI’s Trust Issues

AI is often framed as a technical problem, more code, faster hardware, bigger compute. But in reality, it’s a complex systems challenge that also requires scientific, ethical, and human-centered thinking. Treating it as just “technology” oversimplifies how it actually works in the world.When we overlook fields like logic, philosophy, or behavioral science, we miss critical insights into system behavior, insights essential for building trust in AI systems.

When I first studied AI, logic-based reasoning was foundational. Some may see it as outdated, but I believe its principles still matter, especially when today’s models produce contradictions that point to deeper design flaws.

Rebuilding Trust in AI Systems Through Transparency and Accountability

Only 2 % of U.S. adults say they fully trust AI can make fair and unbiased decisions, and 60 % express some level of distrust, underscoring the urgent need for transparency and accountability. That means prioritizing transparency not only in how models are built, but in how their decisions are made and validated.

I believe explainability is essential. Stakeholders must be able to trace decisions, identify sources of bias, and assess the reliability of outputs. This applies across sectors, whether it’s an algorithm recommending healthcare treatments or one used in public safety systems. Accountability is equally critical. AI systems should include safeguards that support human oversight, especially in high-stakes applications. Agencies must have access to tools that facilitate performance audits, risk assessments, and independent verification and validation.

Transdisciplinary Collaboration: The Key to Responsible AI

Trustworthy AI doesn’t come from one field alone. It depends on collaboration across engineering, policy, economics, and behavioral science. That’s the essence of transdisciplinary work, not just combining tools, but creating new frameworks that bridge disciplines and reflect how complex systems operate in the real world.

From what I’ve seen, this level of collaboration isn’t optional. AI must be shaped by both technical experts and the communities it affects. Only by expanding who’s involved in the design process can we address bias, ensure fairness, and build trust in AI systems that can endure over time.

Moving From Oversight to Confidence: A Call to Action for Leaders

Organizations face growing pressure to deploy AI quickly, but speed can’t come at the cost of trust. Only 54 % of U.S. consumers believe their organizations have responsible AI policies, and 1 in 4 think none exist, showing why oversight alone isn’t enough. Responsible AI must become a strategic priority.

From my experience, here are five practical steps leaders can take:

1. Set clear governance policies. Define how AI is used, reviewed, and evaluated.
2. Ensure model transparency. Use documentation and tools that explain decisions.
3. Build interdisciplinary teams. Include stakeholders, applied scientists, ethicists, and behavioral experts.
4. Conduct independent audits. Test regularly for fairness, drift, and risk.
5. Involve stakeholders. Create feedback loops with those impacted by AI.

These actions help shift AI from a compliance checkbox to a strategic advantage rooted in trust.

Rebuilding Trust Starts with Responsible Action

Rebuilding trust in AI systems takes more than good intentions or cutting-edge technology, it demands sustained effort, clear accountability, and thoughtful collaboration across disciplines. Trust isn’t built overnight. It’s the result of transparent design, rigorous evaluation, and shared responsibility from all parties involved.

At KaDSci, we help organizations energize their data and optimize AI performance, ensuring your models operate with precision and reliability. Reach out to explore how we can help guide your AI initiatives with clarity and purpose.