AI ethics: the imperative for data inclusion and human-centric design

Delivered by Mr. Dixon Siu, the presentation on the importance of AI governance commenced by highlighting the immense potential of AI, demonstrating its innovative applications across multiple critical domains, including health care, education, and agriculture. However, he immediately tempered this optimism with a critical warning: if the development and deployment of AI technologies are left unchecked, they inevitably introduce severe risks such as embedded bias, extensive surveillance, and a fundamental lack of public trust. He underscored that the benefits of AI cannot be fully realized unless these ethical challenges are proactively and systematically managed, making governance an urgent prerequisite for innovation.

Mr. Siu then detailed exactly why a robust framework of AI and data ethics matters, emphasizing the core principles necessary for responsible technology: ensuring transparency, fairness, privacy and consent, and accountability. He vividly demonstrated the consequences of failing to guarantee these ethics, showing how neglect could quickly lead to the exploitation of human rights. A particularly powerful example cited was in the medical domain: many AI-driven medical checkup models are trained on data that do not possess symptoms usually manifested in rural patients, or lack sufficient representation from varied geographic or socioeconomic demographics. This crucial lack of data representation means the models are inherently unreliable for significant portions of the population, leading to misdiagnosis and perpetuating deep-seated healthcare inequality.

To effectively combat this pervasive problem of skewed representation, Mr. Siu introduced the critical concept of sex-disaggregated data (SDD). He defined SDD as data or information that is explicitly separated by gender or sex, displaying results for both men and women individually instead of pooling everyone into the same generalized result and category.

The benefits of this approach are manifold and immediately practical. SDD helps AI systems to detect bias early, fundamentally ensure fairness in outcomes, improve accuracy across various user groups, and allows policymakers to see the real picture, whether that is women’s true representation in the job market or understanding how healthcare works differently for different groups. By seeing the data separated, it becomes impossible to ignore where disparities lie.

Mr. Siu asserted that for AI to be equitable and accurate, the industry must fundamentally change its approach: when designing AI models, we need to make sex-disaggregated data a default instead of an exception in the data collection and modeling process.

Beyond the technical aspects of data, the presentation stressed that AI needs to be human-centric. This approach ensures that technology development is guided by human values, well-being, and societal prosperity rather than pure technological capability or commercial gain.

He briefly contrasted how major global players currently approach this challenge, noting that the EU, China, and the US have different orientations when designing AI models and tech innovation, reflecting varied national philosophies regarding regulation, state control, and market freedom. The focus then shifted to the national context, specifically Cambodia’s National AI Strategy (2025-2030). Mr. Siu explained that the entire strategic document and its pillars, which aim for inclusive national development, can only be robustly supported through the immediate implementation of credible and ethical AI governance.

The presentation concluded with a guiding principle for the future of the technology: AI’s true purpose should be to amplify human values and not replace them, encouraging the audience to see AI as a powerful tool for societal betterment, provided that ethics remain firmly at the steering wheel of development.