I Love the focus on ethics and building trust in AI design, crucial for user confidence and system efficiency! đź‘Ť
AI Design Compass (https://lnkd.in/ebU75PeQ), a comprehensive guide with design consideration to keep in mind when designing AI interfaces — by considering UX strategy, happy and unhappy paths, pitfalls, risks, ethics, feedback and building trust. By Vincent Koc, JP.
From all the design patterns listed in the guide, probably the most significant one is around ethics and building trust. Technically, it’s remarkably challenging for an AI engine to return meaningful results with a high level of accuracy (>95%). It’s also very difficult for AI to understand that it doesn’t know something, and “admit” it.
On the other hand, typing and re-typing text prompts for AI isn’t a great user experience. It’s slow, it’s repetitive, it’s time-consuming, and highly inefficient. Yet AI systems need very specific, highly detailed questions to provide specific, highly relevant answers.
We could ask AI to generate AI prompts for itself. For that, we need to provide users with a rich set of templates and “building blocks” to communicate their intent, then ask AI to generate a variety of scenarios, and then verify relevant ones and dismiss the rest.
For that, we need an efficient way to help users gradually fine-tune AI outcome, ideally without using a text interface at all. For example:
– Allow users to adjust the temperature of output with knobs,
– Cluster/cache AI responses to avoid expensive calculations,
– Allow users to ask for more context to highlight some areas,
– Suggest style lenses (Concrete ↔ Abstract, Lengthy ↔ Short),
– Suggest scopes to limit output to a level of detail or expertise,
– Allow users to scope their queries to a domain, time-frame,
– Restrain AI to provide sources for each conclusion or insight,
– Suggest relevant directions via related AI queries,
– Add structure with chapters, segments to navigate data faster,
– Suggest specific presets and templates to boost efficiency,
– Help users make sense of data by clustering or summarizing it.
AI interfaces can go way beyond text prompts. Once we find a way to surface fine-tuning for AI into our interfaces, we should see user’s efficiency increasing, to the point that it might feel almost magical.
But most importantly: as we give users more control, we also pave a way to build trust into the system and the results it delivers.
✤ Useful resources
AI Interaction Patterns, by Emily Campbell
https://www.shapeof.ai
Design Patterns For Building User’s Trust, via Sarah Gold
https://lnkd.in/etZ7mm2Y
Language Model Sketchbook, by Maggie Appleton
https://lnkd.in/eXfxFk9w
Style Lenses For AI, by Amelia Wattenberger
https://lnkd.in/e-SJis23
AI Ă— Design Toolkit (Worksheets + PDF), by Nadia Piet
https://lnkd.in/eJEkCu2p
AI Design Patterns, by Vitaly Friedman
https://lnkd.in/efzFXbaZ
#ux #design #ai
AI Design Compass, Design Patterns