To overcome this articulation barrier, the author proposes detailed qualitative studies of users with varying literacy skills using AI tools for real business tasks. The insights from these studies could inform the design of more user-friendly interfaces. The author also suggests that successful AI user interfaces may need to combine elements of intent-based outcome specification with aspects of the graphical user interface from the previous command-driven paradigm.
Key takeaways:
- Generative AI systems like ChatGPT rely on users' ability to articulate their needs in written prose, which poses a challenge for half of the population in rich countries who are classified as low-literacy users.
- Articulating needs in writing is difficult even for high literacy users, as evidenced by the prevalence of 'prompt engineers' who specialize in writing text to elicit desired outcomes from AI.
- Improving AI usability requires detailed qualitative studies of users with different literacy skills, and the development of user interface designs informed by this research.
- Adult literacy research findings indicate that half the populations in rich countries are poor readers, which suggests a significant barrier to using advanced generative AI systems.