Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors
About the Author
About the Author
About the Author

Values AI: Bridging the Trust Deficit in Institutions, Business, and Ourselves

Being Human for a Better Tomorrow in the Age of AI

  • June 20, 2025
  • 50th Edition
CivilizationalValues AI Bridging the Trust Deficit in Institutions Business and Ourselves

More Than a Trade War, We’re Fighting a Trust War and a Crisis of Grievance

As AI rewires society, a deeper crisis simmers beneath the surface of geopolitical and economic tensions. It isn’t a battle over tariffs or market share, but a much more fundamental conflict: a global trust deficit that is corroding our most vital institutions, destabilizing markets and a crisis of grievance that is shaking our confidence in the future. 

The diagnosis is stark. The 2025 Edelman Trust Barometer, which, in its 25th year, finds trust in a state of emergency. A generation of institutional failures has left citizens feeling left behind, anxious, and deeply distrustful.

Elections are failing to restore faith; in fact, 11 out of 13 countries that held national elections last year saw trust decrease. Fears over job security—whipped up by globalization, economic pressures, and automation—have intensified.

The result is a deepening “trust penalty” that impacts every facet of society:

  • Leaders Are Not Believed: Distrust in government leaders, business leaders, and journalists has surged by over 10 points in just the last few years.
  • Employers Are Losing Ground: For the first time, global employer trust saw a significant 3-point decline, a worrying trend for the one institution people have traditionally relied upon.
  • Media Is Misaligned: A majority now believe news organizations are more concerned with attracting a large audience or supporting an ideology than informing the public.
  • Grievance Infects Everything: Those with a high sense of grievance are profoundly more distrustful of all institutions. This cynicism extends to technology, with trust in AI dropping a further 10 points among this group.

Paradoxically, this grievance demands more action from institutions like business, not less.

But how can they act effectively when the very foundation of belief has crumbled? To rebuild, we need a new AI architecture—one designed not just for intelligence, but for integrity.

But how can they act effectively when the very foundation of belief has crumbled? To rebuild, we need a new AI architecture—one designed not just for intelligence, but for integrity.

Values AI: A Dual-Engine AI Architecture for Trust

The crisis of trust cannot be solved with more of the same data-driven intelligence that, in part, fuels our anxiety. The solution requires a new kind of AI, one designed not just to process facts but to weigh principles. Let’s call it Values AI.

This is not a single technology but a new dual engine architecture, an integrated system that functions like the human brain’s two hemispheres working in concert:

The "Right Brain" AI (The Generative Engine)

This is the AI we know today—the powerful, pattern-recognising, data-driven Large Language Models (LLMs). They are masters of processing vast information, synthesising content, and identifying correlations. They answer “what is” and “what could be.”

The "Left Brain" AI (The Validation Engine)

This is the innovation. A new class of AI engine trained not on the open internet, but a logic engine built on a curated corpus of foundational logic and principles like Society of Professional Journalists (SPJ), UN Global Compact principles, Universal Declaration of Human Rights, and common code of values, morals and ethics.

Its purpose is not to create information, but to validate it. It acts as a check and balance, asking “is this true?” and “is this right?”

This dual-engine model provides a mechanism for generating a quantifiable, transparent “Trust Score”—a FICO score for integrity.

What Does This Actually Mean in Practice?

For Media Integrity:

Imagine a “Truth in Media” score embedded in your browser. As the Right Brain AI engine surfaces an article, it passes the response to the Left Brain AI engine autonomously assesses it against journalistic principles—verifying sources, community notes and flagging unsourced claims. The resulting score gives citizens an immediate, transparent measure of an article’s credibility.

For Employer Trust:

An internal “Values Agent” could analyse a company’s decisions (e.g., restructuring announcements, policy changes) against its stated public values and employee sentiment data, providing leadership with an unvarnished “Say-Do” gap score, creating a powerful incentive for authentic action.

For Government Accountability:

A civic “Pledge to Progress” agent could track a government’s budget commitments against tangible project outcomes, providing citizens with a public dashboard of how their tax dollars are delivering results, turning opaque processes into transparent accountability.

For Ourselves:

As explored in 30th edition Mindvista editon on in the “Inner Watchman” and using AI for self intelligence, this model offers a technological ally as a dual engine AI thatcan be an Inner Watchman. Our personal AI could help us hold ethical, value-based conversations with ourselves, acting as a Socratic partner to ensure our actions align with our principles.

Adding Values to AI is Real and Not a Hallucination

While the architecture of Values AI is new, there is also recognition in AI tech world for incorporating values in AI.

Anthropic’s Constitutional AI and our “Values AI” concept share a similar noble goal—aligning AI with human principles. Constitutional AI (CAI) is a method for training AI models, particularly large language models (LLMs), to align with human values and ethical principles. It uses a predefined set of rules or principles, often referred to as a “constitution,” to guide the AI’s self-evaluation and revision of its responses.

However Values AI architectural and philosophical approaches are fundamentally different. (See Side bar on Constitutional AI.)

Challenges in Value AI

This dual-engine architecture is a novel idea that addresses a critical gap between AI’s power and its accountability. However, its implementation faces significant hurdles:

  • The Oracle Problem: Who decides the "ground truth"? A "Left Brain" AI is only as good as the principles it's trained on. We hace pointed to corpus of content on ethics, values and morals but there is no authority to define the universal values for this validation engine.
  • The Implementation Reality: Today's right brained AI still struggles to reliably fact-check even simple claims, let alone evaluate complex ethical positions in a nuanced way. The technical gap between a dual engine vision and current capabilities is substantial.

Despite these challenges, the idea has profound merit. Three years ago, few could have imagined the current power of generative AI. The pace of improvement is exponential. Now is the time to propose a fork in this development—to insist that alongside the race for greater intelligence, we must begin an urgent parallel effort to build engines of validation.

Implementing Values AI with Agentic AI

The practical vehicle for this “Left Brain” validation engine of Value AI is the next wave of technology: Agentic AI.

So far, all conversation and use cases around AI agents has been focused on their potential for workflow automation—a more efficient way to book holidays or manage an event. Using Agents for automating processes misses a true transformative potential to add value to humans.

We must reimagine agents not merely as tools of personal productivity, but as instruments for building trust between humans.

Imagine autonomous agents, commissioned by independent bodies or stakeholder groups, with the mandate to serve as trust auditors.

  • A 'Climate Pledge Agent' could access publicly reported emissions data, cross-reference with satellite monitoring services, and flag discrepancies against sustainability commitments—operating within existing transparency frameworks like CDP reporting
  • A "Hiring Equity Agent" could analyze a company's recruitment and promotion data over time to detect systemic bias, reporting its findings directly to a board's audit committee.

This represents a radical shift. The agent is no longer just an assistant to an individual user; it becomes a persistent, autonomous guardian of stakeholder interests.
It’s an AI that works for a ’system’, but for the collective “us.”

Takeaways

For AI Technology Companies:

Pioneer the dual-engine architecture. The future isn’t just about making AI smarter, but making it wiser. The development of “Values AI”—a validation engine to complement the generative engine—is the next great frontier.

For Businesses:

Move beyond efficiency. Deploy agentic AI in pilot projects designed to build stakeholder trust. Use cases like transparent supply chain tracking, verifiable ESG reporting, or internal “Say-Do” accountability tools can turn trust from a liability into a competitive advantage.

For Media Organisations:

Don’t just report on the trust deficit—help solve it. Experiment with developing and publishing “Trust Scores” for sources and content. Make verifiable integrity a new standard of journalistic excellence.

For Individuals:

Empower your “Inner Watchman” with an AI ally. Use AI not just for answers, but to ask better questions about your own actions and their alignment with your values, fostering a new level of conscientious living.

Conclusion

The urgency of this moment cannot be overstated.
As noted in the Edelman 2025 Global Trust Barometer reports that pessimism is setting in, with less than one in five people in developed countries believing the next generation will be better off. Globally, that number is also grim one in three. This is the bitter fruit of a world that has lost faith in its institutions and its future.

Yet, the Edelman data reveals a powerful antidote: When trust increases, economic optimism overpowers grievance.

Restoring trust is therefore the most critical task of our time. We cannot rely on the old institutional playbooks. We must innovate.
Values AI is a proposition for just that—a human-designed system to cultivate and reward our most vital human values: integrity, accountability, and ultimately, trust.

Can we use AI for media to put facts first? Can we use AI to hold businesses, instituitions and goverment to account? Can tech enable a future for all and not one at the cost of another?

Your take?

A positive answer can not just better our trust scores but makes us More Human in the Age of AI.

As we saw in the last three editions, AI Allyship gives us the power to do more. AI as a Connection Catalyst helps us relate better. Civilisational AI protects our collective wisdom.

And Values AI provides the ethical compass to guide it all. Together, they form a coherent vision for not just being human, but increasing our humanity in the Age of AI.

Will AI technology and implementation focus on building trust? What else can we do to bridge the trust deficit? Convert crisis of grievance to optimism for future for all?

Learn, explore and inspire!

Sidebar: Anthropic’s Constitutional AI – A Parallel Path to Values AI

The quest for value-aligned AI is not new, and one of the most significant efforts in this space is Anthropic’s “Constitutional AI” (CAI), the training methodology behind its model, Claude.

Instead of a separate validation engine, CAI aims to build ethical guardrails directly into a single generative model. The process is a form of self-correction: the AI is prompted to generate responses to a variety of requests, including harmful ones. It is then asked to critique its own responses based on a “constitution”—a set of guiding principles (e.g., “avoid toxic or discriminatory outputs”). Finally, the AI fine-tunes itself based on its own critiques.

While the goal is similar to Values AI, the approach differs fundamentally on three key points:

  • Single-Engine Enhancement vs. Dual-Engine Architecture: CAI works by refining a single model. Our proposed Values AI architecture is a dual-engine system, explicitly separating the generative "Right Brain" from the validation "Left Brain."
  • Self-Correction vs. Explicit Training: CAI learns values through an iterative process of self-critique based on scenarios and human-written principles. The Values AI "Left Brain" is envisioned as being explicitly trained on a foundational corpus of established legal, ethical, and civic codes.
  • Proprietary vs. Universal: CAI is a groundbreaking technique specific to Anthropic's models. Values AI is proposed as a universal, cross-platform architecture that could theoretically validate outputs from any generative model.

Constitutional AI represents a vital and pioneering step toward creating safer individual models. Our dual-engine Values AI architecture, however, is proposed to address the broader societal challenge: creating an independent, systemic layer of validation needed to rebuild trust across the entire information ecosystem.

Best wishes

 

Reference:

https://www.edelman.com/trust/2025/trust-barometer

“AI is a language. Treat it like one: practice, iterate, and mind your grammar prompts, assumptions, and verification.”

II. Questioning / Asking

Good conversations flow from well‑sequenced questions—topical, simple, coherent, cohesive.
LLM Conversation Example 1
Q: What are empirical judgments?

A: Empirical judgments are based on observation, experience, or experimentation.
Q: What are moral judgments?

A: Moral judgments are based on ethical principles and values.

IV. Judgment

“The structure of a language affects its speakers’ worldview and cognition.”
— Henry Hazlitt
“The art of questioning is the source of all knowledge.”
— Thomas Berger
Share: