Values-Based AI Governance

Definition

Values-based AI governance is the practice of defining AI use through explicit institutional values, moral commitments, and social-impact boundaries rather than only technical capability or efficiency.

Current Synthesis

In banking, Sajit Channa of Beneficial State Bank frames AI governance as a values and morals problem as well as a technical one. Banks need AI guidelines that address labor markets, privacy, access to credit, and discrimination before efficiency becomes the only deployment test. Juan Verde and Shekhar Natarajan broaden the same principle beyond banking by comparing AI with globalization: institutions lose legitimacy when aggregate efficiency excludes worker dignity, community effects, agency, and transition costs.

This does not reject efficiency. Its stronger claim is that gains become institutionally usable only when an organization can say what values constrain deployment, which harms it is trying to avoid, how affected people retain agency, and how responsible use can widen rather than narrow opportunity.

Key Claims

• AI guidelines should encode what an institution is willing and unwilling to automate.
• Labor-market impact, privacy, access to credit, and discrimination should be explicit governance categories.
• Values-based governance is compatible with business goals when nondiscrimination expands reachable borrower communities.
• Internal values do not replace regulation, vendor oversight, or technical validation; they guide what the institution asks those systems to protect.
• In banking, values-based governance has higher stakes because AI can influence credit access and local economic opportunity.
• Outside banking, labor displacement, community disruption, human dignity, and transition preparation are explicit governance categories rather than after-the-fact externalities.
• Candid internal advisers are part of the control environment when they can challenge an efficiency-only business case before deployment.

Evidence

• Guideline claim: AI in banking: the good, the bad, and the efficient has Sajit Channa argue that banks need AI guidelines grounded in values and morals.
• Impact categories: AI in banking: the good, the bad, and the efficient says Channa names labor markets, privacy, and access to credit.
• Access-plus-business claim: AI in banking: the good, the bad, and the efficient says avoiding discrimination can help banks make more loans by reaching broader communities.
• Dignity-and-transition claim: Juan Verde on Politics, Purpose, Leadership, Sustainability and the Future | Shekhar Natarajan compares AI with globalization’s worker and community costs and calls for values-aware decisions before efficiency dominates.
• Leadership mechanism: Juan Verde on Politics, Purpose, Leadership, Sustainability and the Future | Shekhar Natarajan argues for advisers willing to ask uncomfortable questions and tell leaders what is actually happening.

Counterevidence & Qualifications

Values language can remain aspirational unless paired with operational controls, audits, vendor requirements, worker-transition plans, and decision accountability. The banking source does not provide Beneficial State Bank’s detailed policy, implementation checklist, or measured outcomes. The Verde interview supplies a broad historical analogy rather than sector-level displacement estimates or a concrete AI governance standard.

What Changed

• Broadened the concept from banking-specific values to labor, dignity, community, agency, and transition effects across sectors.
• Added globalization as a cautionary analogy for efficiency gains that externalize concentrated social costs.
• Added candid internal advice as a leadership mechanism for surfacing omitted harms.

Related Concepts

• AI Governance And Compliance - broader governance system that values-based rules must connect to.
• AI Credit Access Bias - discrimination and access risk that values-based governance should address.
• Community Bank AI Adoption - adoption context for the source’s banking example.
• Third-Party AI Vendor Oversight - vendor layer that internal values need to constrain.
• Human Judgment Under AI - accountability boundary where people define and own final institutional judgment.
• AI Alignment Governance - adjacent institutional-values frame for aligning AI systems and organizations.
• Efficiency Without Dignity - failure mode that values-based governance is intended to detect before deployment.
• Constructive Dissent / 建设性异议 - organizational feedback practice that keeps uncomfortable evidence available to decision-makers.