Generative AI-Enhanced Knowledge Engineering

In the last few years, many have suggested that generative AI makes human-mediated knowledge engineering of encoded data, knowledge graphs, value sets, and clinical decision support rules obsolete. Even the leading generative AI tools can still hallucinate. They don’t necessarily know what they don’t know. At Elimu, our knowledge engineers are utilizing generative AI to make the process much more efficient and effective and while doing so, uncovering the risks of allowing these solutions to build content without human oversight.

The Promise of Generative AI in Knowledge Engineering

Generative AI offers a compelling vision for the future of knowledge engineering. Its ability to process vast amounts of unstructured data, identify patterns, and generate new content holds immense potential for accelerating and enriching the creation of structured knowledge.

• Accelerated Data Extraction, Encoding and Classification: Generative AI can rapidly extract relevant information from diverse sources such as existing structured clinical data dictionaries, scientific literature, clinical guidelines or regulatory documents. This significantly reduces the manual effort traditionally required to identify and encode data for knowledge graphs or value sets.
• Enhanced Knowledge Graph Construction: By analyzing relationships between entities in guidelines and literature, generative AI tools can draft knowledge graphs or propose new connections to enrich existing knowledge graphs. This can lead to more comprehensive and nuanced representations of complex domains like clinical genomics.
• Automated Rule Generation and Validation: In domains like clinical decision support, generative AI could assist in drafting initial versions of rules or even validating existing ones by identifying potential inconsistencies or gaps. This can streamline the development and maintenance of critical decision support systems.
• Improved Efficiency and Scalability: The sheer processing power of generative AI allows for knowledge engineering tasks to be performed at a speed and scale previously unimaginable. This means faster development cycles, continuous updates, and the ability to tackle more ambitious knowledge representation projects.

The Indispensable Role of Human Curation: Mitigating the Risks

While the opportunities are transformative, the risks associated with unbridled generative AI in knowledge engineering are significant and benefit from expert human oversight.

• Combating Hallucinations and Inaccuracies: The most pressing risk is the propensity of generative AI to “hallucinate” — generating plausible-sounding but factually incorrect information. In critical domains like healthcare, inaccurate knowledge can have severe, even life-threatening, consequences. Human knowledge engineers are essential for verifying the accuracy of AI-generated content.
• Addressing “Unknown Unknowns”: Generative AI operates based on the data it has been trained on. It doesn’t inherently understand what it doesn’t know or the limitations of its own knowledge base. Human experts, with their domain expertise and critical thinking skills, can identify gaps in the AI’s understanding.
• Ensuring Contextual Understanding and Nuance: Knowledge engineering often requires a deep understanding of context, subtle nuances, and implicit relationships that generative AI may struggle to fully grasp. Human knowledge engineers bring the ability to interpret complex information, resolve ambiguities, and ensure the encoded knowledge accurately reflects the intended meaning.
• Maintaining Ethical and Bias Standards: AI models can inadvertently perpetuate biases present in their training data. Human oversight is crucial to identify and mitigate these biases in the generated knowledge, ensuring fairness, equity, and ethical considerations are embedded in the resulting systems.
• Strategic Direction and Problem Formulation: While AI can generate content, humans are still needed to define the problems to be solved, formulate the knowledge engineering goals, and strategically guide the generative AI tool’s approach.

A Synergistic Future

At Elimu, we believe the future of knowledge engineering lies not in the replacement of human expertise by AI, but in a powerful synergy between the two. Generative AI serves as an incredibly potent tool, amplifying the capabilities of human knowledge engineers. It frees them from tedious, repetitive tasks, allowing them to focus on higher-level activities such as:

• Validation and Vetting: Critically evaluating AI-generated content for accuracy, completeness, and contextual relevance.
• Gap Analysis and Refinement: Identifying areas where the AI’s knowledge is incomplete or inaccurate and guiding its further development.
• Complex Problem Solving: Applying human judgment and domain expertise to address highly complex or ambiguous knowledge challenges.
• Ethical Oversight: Ensuring the responsible and unbiased application of AI in knowledge creation.

By embracing this collaborative approach, we can harness the immense power of generative AI to build more comprehensive, efficient, and accurate clinical decision support systems, all while safeguarding against its inherent limitations through the invaluable contribution of human expertise and critical oversight. The marriage of human curation with generative AI’s comprehensiveness and efficiency is not just an opportunity; it’s the responsible path forward for the evolution of knowledge engineering.