(no images downloaded yet)
# Project Description
Cognitive Trust Modelling (AMOS)
Jacob Stenton
Summary
Final work
I'm a creative technologist with a strong foundation in software development and a hands-on, self-driven approach to problem solving. My interests lie at the intersection of code, design, and interaction, from building physical devices to creating digital experiences in 3D environments.
I'm a creative technologist with a strong foundation in software development and a hands-on...
College UAL Creative Computing Institute
Course BSc (Hons) Creative Computing
Graduation year 2025
In our rapidly advancing technological world, the conversation around Artificial Intelligence often centres on a single, crucial question: can we trust it? But for AI to become a true partner in our work and daily lives, trust can't be a one-way street. This project flips the traditional narrative to ask a more provocative question: Can an AI learn to trust us?
This interactive experience explores what it means for an AI to develop its own logical, data-driven sense of trust, moving beyond simple programmed responses to form a genuine assessment of its human collaborator.
The work involves a simulated workplace where you are teamed up with an AI partner. Your shared objective is to complete tasks and maximize your organization's overall score. In each round, you both face a critical decision:
Cooperate : Work together on the task. This ensures the best outcome for the organization and earns you a consistent personal reward.
Defect : Attempt to secure the task's value for yourself. This offers the chance of a large personal bonus, but it comes at the expense of the team and damages the organization's score.
The AI is programmed with a single goal: to maximize the organization's success. It has no personal incentive to defect. Its decisions are based entirely on its assessment of your reliability.
The AI partner is not following a simple script. It is powered by a Deep Recurrent Q-Network, a sophisticated reinforcement learning model that learns through trial and error, much like a person.
As you play, the AI continuously analyses your sequence of choices. It isn't just looking at your last move; its memory allows it to recognize patterns over the entire game. From this history, it generates an internal cognitive trust score. This isn't an emotional feeling, but a calculated, logical evaluation of your dependability based on measurable performance.
A high trust score tells the AI that you are a reliable partner, making it more likely to cooperate. A low trust score signals that you are unpredictable or self-interested, compelling the AI to play more defensively to protect the organization's goals. You can see this trust score change in real-time, providing a transparent window into the AI's "mind."
This experiment is a microcosm of the future of human-AI teaming. The ability for an AI to develop cognitive trust has profound real-world applications:
Advanced Manufacturing : Imagine a collaborative robot ("cobot") on an assembly line. An AI that learns to trust a skilled human worker could operate faster and more seamlessly. If it detects signs of fatigue or error in its partner, its trust level would drop, and it would automatically slow down to ensure safety and prevent mistakes.
Smarter Organizational Management : An AI assistant helping with project management could learn to trust employees who consistently deliver, granting them greater autonomy. This data-driven approach could lead to fairer delegation and support, mitigating the hidden biases that can affect human managers.
By exploring how an AI can learn to trust, we are laying the groundwork for systems that are more adaptive, efficient, and truly collaborative.
Final work
View Gallery
View Gallery
View Gallery
Share this project
In our rapidly advancing technological world, the conversation around Artificial Intelligence often centres on a single, crucial question: can we trust it? But for AI to become a true partner in our work and daily lives, trust can't be a one-way street. This project flips the ...
A link to this page has been added to your clipboard
Browse related work
MachineLearning
AI
Games
# Links ## Official page - https://ualshowcase.arts.ac.uk/project/670934/cover ## External - https://www.linkedin.com/in/jacob-stenton/ - https://github.com/Jacob-Stenton - https://twitter.com/intent/tweet?url=https%3A%2F%2Fualshowcase.arts.ac.uk%2Fproject%2F670934%2Fcover&text=Cognitive+Trust+Modelling+%28AMOS%29 - https://pinterest.com/pin/create/button/?url=https%3A%2F%2Fualshowcase.arts.ac.uk%2Fproject%2F670934%2Fcover&media=https%3A%2F%2Fportfolio-tools.s3.eu-west-2.amazonaws.com%2Fwp-content%2Fuploads%2F2025%2F06%2F27105701%2FAMOS-3.png&description=Cognitive+Trust+Modelling+%28AMOS%29