Too many articles have been published recently lamenting how "AI is making us dumb". Such an empowering technology should not disempower its users. This rhetoric should challenge for engineers and designers to design better interfaces that leave users feeling in control and agentic.
I set out to reimagine an interface in which users' productivity could benefit from AI tools, without compromising users' agency. I additionally wanted to create affordances for learning opportunities. For this V0, I limited my design to LLM chat interfaces and focused on the task of editing/refining/improving textual information (i.e. asking an LLM to debug your code or refine your written paragraph).
My prototype does two things:
1. the model makes the minimally needed edits to user input
2. the edits are clearly displayed to the user
These features make it such that the LLM is actually building on the user, creating opportunities for the user to learn how their own work could be improved. And with edits being so clearly highlighted, users easily maintain control over what to accept from the LLM.
Interested in hearing people's thoughts on this more collaborative paradigm of human-machine interaction!
manooshree•6mo ago
Youtube demo at: https://www.youtube.com/watch?v=GzyrazM_dqY&ab_channel=Manoo...
Too many articles have been published recently lamenting how "AI is making us dumb". Such an empowering technology should not disempower its users. This rhetoric should challenge for engineers and designers to design better interfaces that leave users feeling in control and agentic.
I set out to reimagine an interface in which users' productivity could benefit from AI tools, without compromising users' agency. I additionally wanted to create affordances for learning opportunities. For this V0, I limited my design to LLM chat interfaces and focused on the task of editing/refining/improving textual information (i.e. asking an LLM to debug your code or refine your written paragraph).
My prototype does two things: 1. the model makes the minimally needed edits to user input 2. the edits are clearly displayed to the user
These features make it such that the LLM is actually building on the user, creating opportunities for the user to learn how their own work could be improved. And with edits being so clearly highlighted, users easily maintain control over what to accept from the LLM.
Interested in hearing people's thoughts on this more collaborative paradigm of human-machine interaction!