Page 6 - AI Vol 2: Risks of AI
P. 6
reveal the underlying logic of its conclusions. system's decisions consider relevant factors or
For example, rather than having an AI model mask other influences. Prompting strategies that
determine which students should be placed in an take a targeted approach to decision making, and
academic intervention program, a more effective require AI systems to explain their logic, empower
prompt would be to ask the AI model to identify human decision makers to make informed
students struggling with specific aspects of the decisions about whether to accept, modify, or
curriculum, such as multiplication, and to explain reject AI recommendations. It also fosters a more
the factors the AI system considered to reach that critical and engaged approach to using AI tools,
conclusion. This process could be repeated over ensuring that these powerful technologies are
various aspects of curriculum, and at that point used responsibly and ethically.
the AI model could be asked to identify students
in need of an academic intervention program and In navigating the complexities of AI, public
to explain the factors the AI considered for each agencies should proactively address the
student. This approach not only provides specific inherent risks of bias through rigorous testing,
insights but also enables human reviewers to human oversight, and strategic prompting. By
understand the basis of the AI's decisions. While implementing these measures, agencies can
the AI system may not always be transparent in harness the potential of AI while upholding
explaining its decisions, this approach allows their commitment to fairness, transparency, and
the human reviewer to assess whether the AI accountability in public service.
6 | VOLUME 2 RISKS OF AI | LOZANOSMITH.COM