As Cognizant's Head of Generative AI, I've been immersed in exploring how generative AI can mirror human cognitive processes, especially in healthcare settings. Our recent collaboration with our client, a National Department of Health, has been a testament to this exploration.
We've crafted an AI system that not only performs tasks but also mirrors the nuanced thought processes of human clinicians. The Open Conversational Design Pattern for Enterprise Generative AI, developed for this project, serves as a blueprint for creating AI that closely aligns with human cognition.
The Open Conversational Design Pattern: Human Analogue Approach
This design pattern is not just about task execution; it's about emulating the layers of human thought and interaction. Here’s how each layer of our design pattern aligns with human cognitive processes:
Intent and human catalyst
In human interaction, understanding intent involves more than just processing words; it requires empathy and context. Our AI's first layer mimics this by dynamically engaging with patients, interpreting their words, tone, and context to accurately grasp their needs, much like a human clinician would.
Information as human synthesis
The transition from intent to information gathering in humans involves synthesising what is heard into actionable knowledge. Our AI system mirrors this by delving into vast medical databases, contextualising patient information against current medical knowledge, akin to a clinician’s research and knowledge application.
Cognition: Human-like analysis and deduction
Just as a human clinician uses their expertise to analyse information and form a diagnosis, our AI uses its 30 specialised personas to collectively analyse data and deduce diagnoses. This mimics the multi-faceted decision-making process found in clinical teams.
Presentation: Reflecting human communication
In human interaction, presentation is key. It’s not just what is said, but how it’s said. Our AI system’s presentation layer is designed to communicate its findings and recommendations in a manner that is clear, empathetic, and actionable, mirroring a clinician’s way of communicating complex information to patients.
Achievements in human-like AI application
Our system's success in the Department of Health project, evaluated by over 100 healthcare professionals, showcased its ability to not just process information, but to do so in a manner strikingly similar to human cognition.
Challenges in emulating human cognition
Emulating the nuances of human cognition is challenging. Language barriers and capturing nuanced intent posed significant hurdles, highlighting the complexity of human thought and the ongoing need to refine AI cognition.
Towards a transparent and human-analogue AI
In advocating for transparent AI systems, we propose a regulatory framework that not only evaluates AI performance but also its ability to parallel human thought processes. Such a framework would assess AI systems for their human-like processing capabilities at each stage of the design pattern.
Our journey with the Department of Health has been more than just about creating an AI system; it's been about creating an AI that thinks and interacts like a human, especially in complex, sensitive fields like healthcare. This endeavour underlines the need for AI systems that not just solve problems but do so by closely emulating the intricate processes of human thought and communication.
