Entrust TQ– Intelligent Evaluation & Verification of Trust in AI

Scales weight the factors involved in entrusting AI, created by the author in DALLE-3

Creating a viable system to quantify trust scoring system for AI & intelligent systems, akin to the levels of autonomous driving or NPS, involves establishing criteria that assess various aspects of entrustment. This system is composed of a few different elements which when combined will give a comprehensive entrustment score. At its core this method is a measurement of assurance or risk, like the FICO score that an organization or system has a certain degree of risk they are untrustworthy based on their score. Proposed Structure:

a) A level of trust scale — the scale will have multiple levels, each reflecting a deeper and more comprehensive integration of trust elements in an organization’s operations.

b) A MET or minimum effective trust — the smallest level of trust required to interact.

c) MMoS or minimum margin of safety — the concept of safety margins or error buffers in AI systems.

Level of Trust Scoring Scale

Level 0: Minimal Trust

• Characteristics: Basic compliance with legal standards, minimal customer engagement.
• Value of Trust: Low, primarily focused on avoiding legal repercussions.
• Trust Risk: High, due to lack of proactive trust-building measures.
• Margin of Trust Buffer: Minimal, relying on basic legal compliance.
• Costs of Lack of Trust:
• High legal and reputational costs.
• Strength of Connection: Weak, with minimal interaction between nodes.
• Degree of Coupling: Low, independent operation of nodes.
• Volatility: High, prone to rapid and unpredictable changes in trust levels.
• Resilience: Low, inability to recover quickly from trust breaches.

Level 1: Basic Trust

• Characteristics: Standard product reliability, basic alignment with marketing messages.
• Value of Trust: Moderate, aiming for customer satisfaction and retention.
• Trust Risk: Moderate, with some vulnerability to market changes.
• Margin of Trust Buffer: Limited, mostly reactive measures.
• Costs of Lack of Trust:
• Moderate impact on sales and customer loyalty.
• Strength of Connection: Moderate, occasional interactions between nodes.
• Degree of Coupling: Moderate, some interdependencies.
• Volatility: Moderate, some fluctuations in trust levels.
• Resilience: Moderate, recovery from trust issues is achievable but not rapid.

Level 2: Established Trust

• Characteristics: Consistent product quality, alignment with branding, and customer-centric policies.
• Value of Trust: High, contributing to strong customer relationships and brand loyalty.
• Trust Risk: Lowered, with proactive measures in place.
• Margin of Trust Buffer: Adequate, with established policies for trust maintenance.
• Costs of Lack of Trust:
• Lower due to established trust measures.
• Strength of Connection: Strong, regular and meaningful interactions.
• Degree of Coupling: High, significant interdependencies.
• Volatility: Lower, more stable trust levels.
• Resilience: High, able to recover quickly from setbacks.

Level 3: Advanced Trust

• Characteristics: High brand reputation, strong alignment between marketing messages and customer experience.
• Value of Trust: Very high, integral to brand identity.
• Trust Risk: Lower, with trust deeply embedded in operations.
• Margin of Trust Buffer: Significant, with robust trust safeguards.
• Costs of Lack of Trust:
• Effectively managed with comprehensive strategies.
• Strength of Connection: Very strong, deep and consistent interactions.
• Degree of Coupling: Very high, closely integrated systems.
• Volatility: Minimal, trust levels remain stable over time.
• Resilience: Very high, quick and effective response to trust issues.

Level 4: Exceptional Trust

• Characteristics: Industry-leading in trust, transparency, and open communication; exceptional product reliability.
• Value of Trust: Paramount, a key market differentiator.
• Trust Risk: Minimal, with trust ingrained in every aspect.
• Margin of Trust Buffer: Extensive, strong emphasis on maintaining trust.
• Costs of Lack of Trust:
• Minimal, strong focus on preemptive strategies.
• Strength of Connection: Exceptional, highly responsive and adaptive interactions.
• Degree of Coupling: Integral, functioning as a cohesive unit.
• Volatility: Very low, trust levels are extremely stable.
• Resilience: Exceptional, capable of handling and recovering from trust issues effortlessly.

Level 5: Ultimate Trust

• Characteristics: Pioneering in trust dynamics, setting industry standards, complete transparency, seamless communication.
• Value of Trust: Central to business ethos, driving innovation and engagement.
• Trust Risk: Virtually non-existent, with trust as a core principle.
• Margin of Trust Buffer: Comprehensive, continuous improvement in practices.
• Costs of Lack of Trust:
• Negligible, strong legacy of trust minimizing any potential costs.
• Strength of Connection: Connections are deeply embedded and mutually supportive, enhancing the functionality and reliability of the entire system.
• Degree of Coupling: Total, with all elements of the system being fully integrated and interdependent. The system functions as a seamless whole, with each part enhancing and supporting the others.
• Volatility: Almost non-existent, indicating an extremely stable and predictable environment. Trust levels remain consistent even in the face of challenges or changes.
• Resilience: Supreme, with the ability to anticipate, adapt to, and recover from any trust-related challenges with remarkable efficiency.

This scoring system enables organizations to assess their current level of trust and identify areas for improvement. By advancing through the levels, an organization can significantly enhance its relationship with stakeholders, improve its market positioning, improve strength of connections, shape the structure of and reduce the costs associated with a lack of trust.

Minimum Effective Trust (MET)

The concept of Minimum Effective Trust (MET) is the minimum level of trust required for an interaction to take place.

MET Scoring Components:

1. Levels of Trust (LoT) Score:

a. Based on the previously developed Levels of Trust Scale (Level 0 to 5).

b. Scored from 0 (Minimal Trust) to 5 (Ultimate Trust).

2. Degree of Trust Required by Industry:

a. Industries rated on a scale of 1 to 5, based on the degree of trust necessary for their operation.

b. Example: Healthcare (5), Financial Services (5), Legal (5), Retail (3), Manufacturing (2).

3. Trust Attributes Scoring:

a. Each attribute (Integrity, Honesty, Perception, Predictability, Reliability, Safety, Uncertainty, Transparency, Security, Privacy) is scored on a scale of 1 to 5.

b. The score reflects how well a business or entity performs in each attribute.

Calculating MET Score:

1) Calculate Average LoT Score:

a) Based on operational practices, customer feedback, and internal policies.

2) Adjust with Industry Trust Requirement:

a) Multiply LoT Score by the Industry Trust Requirement factor.

3) Average Trust Attributes Score:

a) Calculate the average score across the 10 trust attributes.

4) Final MET Score:

a) Combine the adjusted LoT Score and the Trust Attributes average score for a final MET score.

Industry Trust requirement Scale, created by the author.

Minimum Margin of Safety (MMoS)

MMoS Scoring Components:

1) Risk Assessment:

a) Evaluate potential risks associated with trust breaches in the specific industry.

b) Scored from 1 (Low Risk) to 5 (High Risk).

2) Safety Measures in Place:

a) Assess the safety measures implemented by the business or entity.

b) Scored based on effectiveness, coverage, and responsiveness.

3) Historical Performance:

a) Analyze past incidents, breaches, or failures in trust and safety.

b) Scored based on frequency, severity, and resolution effectiveness.

i) Red = frequent, significant, and ineffective resolution, score = -10

ii) Yellow = frequent, moderate, and ineffective resolution, score = -5

iii) Green = minimal, moderate, and semi-resolution, score = 0

iv) Bronze occasional, minimal, and ineffective resolution, score = +5

v) Silver = infrequent, minimal, and effective resolution, score = +10

vi) Gold = infrequent, minimal, and preventative, score = +15

Calculating MMoS Score:

1) Risk and Safety Balance:

a) Evaluate how well safety measures counterbalance the assessed risks.

b) Score from 1 (Low Risk) to 5 (High Risk).

2) Adjustment for Historical Performance:

a) Modify the balance score based on historical performance data.

b) Score Red to Gold

3) Final MMoS Score:

a) Risk * Historic performance = MMoS

b) Present a comprehensive score that reflects the current margin of safety.

Final Trust IQ Score

TIQ = MET (LoT * Industry Trust + Trust attribute score) + MMoS

*The higher the number the better

*Note also considered calling this trust scoring system — FAITH = Framework for Assessing Integrity and Trust in AI and Intelligent Systems

Application:

1. Example: Precision Medicine Clinic startup which offers data driven healthcare in person or virtually. As a start up but one whose founding doctors have an established brand; they start with an LoT =2, Industry score = 5, Attribute score = 38, Risk score = 3 (startup), & color bronze.
2. TIQ = (2*5)+38 + (3*5)
3. TIQ = 48+15
4. TIQ = 63
5. They provide a standardized, quantifiable way to evaluate trust and safety, aiding in decision-making and policy formulation.
6. These scores can be used by consumers, businesses, NGO’s, organizations, and government/public officials to assess and compare the trustworthiness and safety levels of different entities.

The LoT, MET, and MMoS concepts combine to create a simple scoring system. This system helps understand and quantify trust and safety in an easy to use risk metric called the Entrust TQ. It is valuable for a wide range of stakeholders in different contexts. When used with the entrust framework, they can create aligned, innovative, and anti-fragile AI and intelligent systems.