Trustworthy and explainable AI – a key to equitable oncology?

https://dailyreporter.esmo.org/spotlight/trustworthy-and-explainable-ai-a-key-to-equitable-oncology

Publish Date: 2026-02-12 04:48:00

Source Domain: dailyreporter.esmo.org

Reliable AI is Essential: For AI-based solutions to be widely adopted in clinical practice, their trustworthiness must be high, and this was highlighted by Professor Alessandra Pedrocchi from Politecnico Milan during her talk at the ESMO Women for Oncology Forum.
Explainable and Trustworthy AI: AI should serve as an additional layer of reasoning, akin to obtaining a second opinion, rather than an automated process. Explainability is crucial to gain clinicians’ trust, which determines AI adoption in healthcare.
Multidisciplinary Collaboration is Key: To effectively engage in oncology, professionals without medical backgrounds need to cultivate a deep understanding of clinical needs and language through close collaboration with medical experts.
Biases in AI Development: Sources of bias in AI development in oncology include unequal data access, biases in algorithms, and how end-users apply the tools. Biases must be actively identified and corrected throughout the development process.
Addressing AI Research Bias: Conducting rigorous exploratory data analysis, stratifying datasets using protected labels, and incorporating external validation data help in identifying and mitigating bias, ensuring robustness and fairness.
Challenges of Small Datasets: In genomics and targeted therapies, small datasets can exacerbate biases as AI algorithms often optimize for the most represented data, making it essential to leverage larger datasets.
The Future of Fair AI: While achieving perfect fairness and inclusivity is yet to come, large-scale data-sharing initiatives, like the European Health Data Space initiative, are crucial towards reducing biases and improving AI solutions to truly reflect real-world cancer complexities.