Trust Erosion Dynamics Model

Overview

This model examines how AI systems contribute to the erosion of trust in institutions, experts, and interpersonal relationships. The 2025 KPMG/University of Melbourne global study found that while 66% of people use AI regularly, only 46% globally are willing to trust AI systems—and trust levels have declined as adoption has increased. Meanwhile, the 2025 Edelman Trust Barometer documented that multi-decade institutional trust erosion has accelerated into what researchers call a "grievance" phenomenon, with 61% of respondents globally reporting moderate to high grievance levels.

The central question this model addresses: How do AI systems erode trust, and why is trust so difficult to rebuild once lost? The answer lies in a fundamental asymmetry: trust erodes 3-10x faster than it builds, and AI technologies dramatically accelerate erosion mechanisms while offering few pathways to restoration.

Trust Erosion Framework

The following diagram illustrates how AI-driven erosion mechanisms cascade through different trust domains, ultimately producing self-reinforcing distrust cycles.

Types of Trust Affected

1. Institutional Trust

• Trust in government, media, science, corporations
• Eroded by: AI-enabled manipulation, deepfakes, surveillance
• Consequences: Governance breakdown, policy resistance

2. Expert Trust

• Trust in professionals, specialists, authorities
• Eroded by: AI competing with experts, AI errors attributed to experts
• Consequences: Ignoring expert advice, dangerous self-reliance

3. Information Trust

• Trust in media, facts, shared reality
• Eroded by: Deepfakes, AI-generated misinformation, authentication failures
• Consequences: Epistemic fragmentation, inability to coordinate

4. Interpersonal Trust

• Trust in other individuals, social relationships
• Eroded by: AI impersonation, synthetic relationships, surveillance
• Consequences: Social atomization, reduced cooperation

5. Technology Trust

• Trust in AI and technology systems themselves
• Eroded by: AI failures, unexpected behaviors, opacity
• Consequences: Resistance to beneficial AI, or paradoxically, excessive trust

Erosion Mechanisms

1. Deepfake Reality Distortion

Mechanism: AI-generated synthetic media makes it impossible to trust visual/audio evidence. Globally, deepfake-related misinformation rose by 245% year-over-year in 2024, with spikes in countries holding major elections.

Process:

• Deepfakes become increasingly convincing—research confirms humans cannot consistently identify AI-generated voices
• Authentic content becomes indistinguishable from fake
• All evidence becomes suspect (the "liar's dividend" allows dismissal of authentic recordings as probable fakes)
• Visual evidence loses probative value in legal and journalistic contexts

Trust Impact:

• Media trust: Severe erosion—Deloitte's 2024 study found 50% of respondents more skeptical of online information than a year ago
• Legal evidence trust: Significant erosion
• Interpersonal trust: Growing concern—68% of those familiar with generative AI report concern about deceptive synthetic content

Current Status: Early-to-mid stage; detection still possible but rapidly declining. While 57% of people believe they could spot a deepfake, research suggests this confidence is misplaced for high-quality synthetic media.

Timeline:

• 2020-2023: Detectable deepfakes, limited impact
• 2024-2026: Near-undetectable deepfakes, significant impact; fraud losses projected to grow from $12.3B (2023) to $40B (2027)
• 2027+: Post-authenticity era (assuming no breakthrough in verification)

2. AI-Enabled Disinformation

Mechanism: AI dramatically scales and personalizes disinformation campaigns.

Process:

• AI generates vast quantities of convincing misinformation
• Personalization makes disinformation more persuasive
• Detection cannot keep pace with generation
• Information environment becomes unreliable

Trust Impact:

• Media trust: Severe erosion
• Platform trust: Moderate erosion
• Peer information trust: Moderate erosion

Current Status: Active and accelerating

3. Surveillance Chilling Effects

Mechanism: Awareness of AI surveillance erodes trust in private communication and institutions.

Process:

• AI enables pervasive surveillance (facial recognition, communications monitoring)
• People assume they are being watched
• Self-censorship and guardedness increase
• Authentic interaction and trust formation impaired

Trust Impact:

• Government trust: Severe erosion (in surveillance states)
• Institutional trust: Moderate erosion
• Interpersonal trust: Moderate erosion

Current Status: Severe in authoritarian contexts, emerging in democracies

4. Expert Displacement

Mechanism: AI competing with and sometimes outperforming human experts undermines expert trust.

Process:

• AI provides faster, sometimes better answers than experts
• But AI also makes confident errors
• Unclear when to trust AI vs. human expert
• Both AI and human expert trust become uncertain

Trust Impact:

• Expert trust: Moderate erosion
• Professional institution trust: Moderate erosion

Current Status: Emerging, accelerating with LLM adoption

5. Authentication Collapse

Mechanism: AI impersonation undermines ability to verify identity and authenticity.

Process:

• AI can impersonate voices, faces, writing styles
• Traditional authentication methods fail
• Impossible to verify identity of remote communications
• Fundamental interpersonal trust undermined

Trust Impact:

• Interpersonal trust: Potentially severe erosion
• Transaction trust: Moderate erosion
• Legal identity trust: Growing concern

Current Status: Early stage but accelerating

Model Parameters

The following table quantifies key parameters in the trust erosion model, drawing on recent survey data and research.

Trust Dynamics Model

Building vs. Eroding Trust

Trust Building:

• Slow, cumulative process
• Requires repeated positive interactions
• Depends on vulnerability and follow-through
• Takes years to build strong trust

Trust Erosion:

• Can be rapid (single betrayal)
• Negative events weighted more than positive
• Cascades through networks (distrust spreads)
• Generalized from specific failures

Asymmetry: Trust erodes faster than it builds. Research on trust asymmetry confirms the adage that "trust goes on horseback and goes back on foot"—negative information has a greater impact on trust levels than equivalent positive information. Estimated 3-10x asymmetry in speed, with the ratio higher for institutional trust than interpersonal trust.

The Trust Cascade

Single Trust Failure
  ↓ (Generalization)
Category Trust Erosion (e.g., distrust one news source → distrust all news)
  ↓ (Expansion)
Institutional Trust Erosion (distrust media → distrust government)
  ↓ (Network Effects)
Social Trust Erosion (nobody can be trusted)
  ↓ (Feedback)
Self-reinforcing distrust (distrust causes behaviors that confirm distrust)

Quantitative Framework

Trust level $T$ decays over time with negative events according to the following dynamics:

$T(t) = T_0 \cdot e^{-\lambda n(t)} \cdot (1 - \gamma C)$

Where:

• $T_0$ = Initial trust level (0 to 1)
• $\lambda$ = Erosion rate per negative event (0.1-0.5 for institutional trust)
• $n(t)$ = Cumulative negative events experienced
• $\gamma$ = Cascade multiplier (0.2-0.8)
• $C$ = Cross-domain contamination factor (0 to 1)

The key insight is that erosion is multiplicative while rebuilding is additive. A single high-profile trust violation (deepfake-enabled fraud, institutional deception revealed) can reduce trust by 20-40%, while rebuilding requires hundreds of positive interactions over years.

Trust Threshold Analysis

Critical Trust Thresholds

Current Trust Levels (US Estimates)

Note: Estimates based on Edelman 2025, Pew Research 2025, and Gallup surveys. Five of the 10 largest global economies (Japan, Germany, UK, US, France) are among the least trusting nations. In contrast, China reports 72% AI trust.

Approach to Critical Thresholds

Scenario Analysis

The trajectory of trust erosion depends on technological developments, regulatory responses, and institutional adaptation. The following scenarios represent plausible futures over the next decade.

The most likely trajectory (55% combined probability) involves continued trust erosion with varying degrees of mitigation, reaching critical thresholds for government and media trust within 5-10 years. The Urban Institute's 2024 analysis notes that trust erosion is "intertwined with broader issues of polarization, gridlock, and social malaise," suggesting that AI-specific interventions alone cannot reverse the trend.

Trust Rebuilding Challenges

Why Trust is Hard to Rebuild

1. Betrayal Trauma

• Trust violations are remembered longer than trust-building
• Emotional weight of betrayal persists
• Risk aversion increases after violation

2. Changed Baseline

• Once trust is lost, default becomes distrust
• Burden of proof shifts to trustee
• Every interaction scrutinized

3. Confirmation Bias

• Distrust looks for evidence of untrustworthiness
• Positive evidence discounted
• Negative evidence amplified

4. Collective Action Problem

• Individual trustworthiness insufficient
• Need systemic change to rebuild institutional trust
• Coordination difficult when trust is low

5. Generational Effects

• Those who experienced trust violation never fully trust
• Younger generations may have higher baseline distrust
• Cultural transmission of distrust

Rebuilding Requirements

Estimated Rebuilding Time:

• Minor trust violation: Months to years
• Moderate violation: Years to decades
• Severe systemic violation: Generations
• Some violations: May be permanent within living memory

Intervention Strategies

Preventing Erosion

1. Authenticity Infrastructure

• Develop robust content provenance systems
• Create identity verification mechanisms
• Invest in deepfake detection and watermarking
• Challenge: Technical arms race, adoption barriers

2. Transparency and Accountability

• Require disclosure of AI use and capabilities
• Implement algorithmic accountability
• Create meaningful oversight mechanisms
• Challenge: Conflicts with business interests

3. Media Literacy and Epistemic Resilience

• Education on information evaluation
• Critical thinking training
• Healthy skepticism without cynicism
• Challenge: Scale, reaching vulnerable populations

Slowing Erosion

4. Platform Responsibility

• Hold platforms accountable for amplifying distrust
• Require moderation of trust-eroding content
• Incentivize trust-building features
• Challenge: Free speech concerns, business models

5. Institutional Reform

• Address legitimate grievances driving distrust
• Increase transparency and responsiveness
• Demonstrate trustworthiness through action
• Challenge: Institutional resistance to change

Rebuilding Trust

6. Long-term Commitment

• Accept that rebuilding takes years/decades
• Consistent trustworthy behavior over time
• No shortcuts to restored trust
• Challenge: Political/business cycles shorter than needed

7. New Trust Mechanisms

• Decentralized verification systems
• Reputation mechanisms
• Community-based trust networks
• Challenge: May not scale, vulnerable to gaming

Model Limitations

1. Cultural Variation

• Trust dynamics vary across cultures
• Baseline trust levels differ
• Model calibrated primarily on Western/US context

2. Measurement Challenges

• Trust difficult to measure precisely
• Survey responses may not reflect behavior
• Different definitions across studies

3. Causation Complexity

• AI is one factor among many eroding trust
• Isolating AI-specific effects difficult
• Political, economic factors also significant

4. Prediction Uncertainty

• Trust behavior in novel situations hard to predict
• Tipping points may exist but are hard to identify
• Future AI capabilities uncertain

5. Rebuilding Understudied

• Less research on rebuilding than erosion
• Historical analogies may not apply
• AI-specific rebuilding strategies unknown

Uncertainty Ranges

Key Insights

1. Asymmetry is fundamental - Trust erodes faster than it builds, making prevention crucial

2. Cascades are dangerous - Trust erosion in one domain spreads to others

3. Thresholds matter - Below certain levels, trust becomes self-reinforcing distrust

4. Rebuilding is generational - Severe trust violations may only heal across generations

5. AI accelerates existing trends - AI amplifies trust erosion mechanisms that existed before

6. Technical solutions insufficient - Rebuilding trust requires social and institutional change, not just technical fixes

Related Models

• Trust Cascade ModelModelTrust Cascade Failure ModelModels institutional trust as a network contagion problem, finding cascades become irreversible below 30-40% trust thresholds and that AI multiplies attack effectiveness 60-5000x while degrading de...Quality: 58/100 - Cascade dynamics in detail
• Epistemic Collapse ThresholdModelEpistemic Collapse Threshold ModelModels epistemic collapse as threshold phenomenon where society loses ability to establish shared facts, estimating 75-80% combined probability of collapse via authentication failure (35-45%), pola...Quality: 50/100 - Information trust failure
• Deepfakes Authentication CrisisModelDeepfakes Authentication Crisis ModelProjects authentication crisis threshold when detection accuracy falls to 50-55% (chance levels): audio by 2026-2027, images 2025-2027, video 2026-2030. Content provenance (C2PA) has 30% probabilit...Quality: 53/100 - Visual evidence trust

Sources

• KPMG/University of Melbourne: Trust, Attitudes and Use of AI: A Global Study 2025 - Survey of 48,000+ people across 47 countries on AI trust and adoption
• 2025 Edelman Trust Barometer - 25th annual survey of 33,000+ respondents on institutional trust
• Pew Research Center: How the US Public and AI Experts View Artificial Intelligence (2025) - Comparative analysis of public vs expert AI attitudes
• Urban Institute: Understanding the Crisis in Institutional Trust (2024) - Academic analysis of trust erosion dynamics
• Frontiers in Psychology: Trust Asymmetry Research (2023) - Research on asymmetric impacts of negative vs positive events on trust
• Deloitte: Deepfake Disruption Report (2025) - Projections on AI fraud losses and deepfake proliferation
• UNESCO: Deepfakes and the Crisis of Knowing - Analysis of synthetic media's impact on shared reality