Information Technology Trust & AI – Knowledge Management

A post on LinkedIn made me decide that it is time to share my thoughts and opinions on this topic.

"In legal tech, AI may produce summaries, yet humans still double-check, increasing workload despite promises of efficiency."

I confess, I have no idea how the current LLM AI technology works. My perspective is purely that of a 'user,' and my view is based solely on opinion.

In a previous blog I mentioned 3 Trust attributes namely Ability, Benevolence and Integrity. As a group these are named factors for perceived Trustworthiness and in my opinion a synonym for Judgement. The trust process continues with the added next step of ‘perceived risk’, taking the risk and evaluating the outcome for this perceived risk.1** Based on this outcome your judgement and behavioral trust is either reinforced or destroyed. In essence, you are making yourself vulnerable for a desirable outcome. These human trust attributes relate to technology trust attributes as well. Group information into 2 categories for a start, functional (Functional aspects of an IT system describe what the system should do. These are specific behaviours, features, and functions that allow the system to perform tasks or operations.) and non-functional (Non-functional aspects describe how the system should perform or the quality attributes of the system. They define constraints, standards, and overall performance rather than specific functions.)

Functional = What the system does

Non-functional = How the system does it

Therefore, it is possible to evaluate any information technology based on human trust attributes:

Competence: The trustor has the skills to produce the outcome.

Benevolence: Demonstrating goodwill, kindness, and genuine care for the other person's welfare, without expecting anything in return.

Predictability: Consistently behaving in a manner that allows the other person to predict and trust in one's actions and intentions.

Integrity: Being genuine and true to oneself in the relationship, fostering trust through honesty and transparency.

Perceived risk: Accepting risk is involved when deciding to trust the trustor.

Vulnerability: Being willing to be vulnerable and share personal thoughts, feelings, and experiences with others, fostering a sense of intimacy and connection.

Reliability: Demonstrating consistency and dependability in actions and behaviors, fulfilling promises and commitments.

Reciprocity: Establishing a mutual exchange of trust and support, where both parties contribute to the relationship's well-being.

Consistency: Maintaining stable and predictable behaviors over time, reinforcing trust and reliability in the relationship.

Boundaries: Respecting and honoring each other's boundaries and privacy, fostering trust and safety within the relationship.

Believe: Believe that the Trustor is trustworthy.

And not to add too much complexity, the two most important knowledge management attributes, namely memory and meaning, are embedded in the information trust attributes.

You can now determine in your own world if the AI information technology you use or want to use is actually according to your business use case. Ask it questions which you already know the answers to as a knowledge worker in your domain expertise. Ask the system the same question 3 times within new sessions (Test for memory and meaning). Give it scores according to your critical thinking. This will give you a benchmark and a sense of whether you trust the information technology. Keep in mind your context and perspective. This also ties into a previous blog: Interoperability – Human, it is time to learn how to talk to the machine. You could also add a column called 'Weight,' where you assign a weighted percentage to indicate the importance of each score in the table below.

Here an example: (Note: This example is not pointing to any LLM technology or AI Technology out there, it is a hypothesis.)

Trustor: The person who places trust in someone else

Trustee: The AI technology that is being trusted

Question: ‘Give me a definition for knowledge management.’ times 3

Trust Attribute	Definition	Human Score	The score the AI gives itself	Functional, non-functional, or both	Definition
Competence	The trustor has the skills to produce the outcome.		9	Functional	Reflects the system’s ability to perform required tasks correctly.
Benevolence	Demonstrating goodwill, kindness, and genuine care for the other person's welfare, without expecting anything in return.		8	Non-Functional	Relates to user-oriented qualities like ethical design, privacy, and user care.
Predictability	Consistently behaving in a manner that allows the other person to predict and trust in one's actions and intentions.		10	Non-Functional	Tied to consistency, reliability, and expected behaviour over time.
Integrity	Being genuine and true to oneself in the relationship, fostering trust through honesty and transparency.		10	Non-Functional	Relates to security, honesty in data handling, and system correctness.
Perceived Risk	Accepting risk is involved when deciding to trust the trustor.		9	Non-Functional	Tied to risk management, security, and error handling.
Vulnerability	Being willing to be vulnerable and share personal thoughts, feelings, and experiences with others, fostering a sense of intimacy and connection.		9	Non-Functional	Systems need to handle vulnerabilities securely; also relates to reliability under failure conditions.
Reliability (Meaning)	Demonstrating consistency and dependability in actions and behaviours, fulfilling promises and commitments.		10	Non-Functional	Directly maps to uptime, fault tolerance, and consistent performance.
Reciprocity	Establishing a mutual exchange of trust and support, where both parties contribute to the relationship's well-being.		8	Functional	Could map to interactive or collaborative features where the system exchanges data or services with users or other systems.
Consistency (Memory)	Maintaining stable and predictable behaviours over time, reinforcing trust and reliability in the relationship.		10	Non-Functional	Tied to predictable outputs, stability, and behaviour under repeated operations.
Boundaries	Respecting and honouring each other's boundaries and privacy, fostering trust and safety within the relationship.		9	Non-Functional	Relates to access control, data privacy, and respecting user-defined limits.
Believe	Believe that the trustor is trustworthy.		9	Both	Users’ trust in the system may rely on functional correctness (it works as expected) and non-functional qualities (secure, reliable, ethical).

For me (in my opinion), the commonly used abbreviation AI (Artificial Intelligence) actually stands for 'Augmented Information.' True AI would need to satisfy both Knowledge Management and Trust attributes to be considered true Artificial Intelligence.

1** Mayer, Roger C., James H. Davis and F. David Schoorman (July 1995). ‘An Integrative Model of Organizational Trust’. In: The Academy of Management Review 20.3, p. 709.