i agree and disagree with you- a trainer may not understand how or what the connections a neural network will make but any product goes through EXTENSIVE QA before being launched. thus, a trainer will see all of the different outcomes any nn will make inside of a product, they see the gamut of differences. a user may not see that.
additionally, i don't think you give UX designers enough credit. i worked alongside engineers, and met frequently with trainers to talk about what kind of results our prototypes were generating. it was up to me to help articulate what the results meant and how those connections were made.
for example- how do you make a long tail result looking different than a short tail one without making it look like a google search? how do you articulate those differences to users so it just doesn't look like a shitty searching algorithm? that is a design problem- articulating those differences visually- not an engineering problem.
Hi all! I'm actually the author of the article- if you have any comments feel free to leave them.
I specifically chose to address this to ux designers bx ml will be the future of product design and what I had noticed, from working at IBM Watson, was a lack of understanding across design and engineering teams how to "explain" or show what was happening in products using ux. This is actually the start of a column for fast co focusing specifically on design for AI and ml.
So how a product is processing information and serving up results has to be articulated to users- maybe that articulation is a visualization or maybe it's something else? It depends upon the product- but this was a call for thinking about how, if your product uses ml, how do you translate that for users in the design of the product? It's advocating that for ML we move away from minimal design.
additionally, i don't think you give UX designers enough credit. i worked alongside engineers, and met frequently with trainers to talk about what kind of results our prototypes were generating. it was up to me to help articulate what the results meant and how those connections were made.
for example- how do you make a long tail result looking different than a short tail one without making it look like a google search? how do you articulate those differences to users so it just doesn't look like a shitty searching algorithm? that is a design problem- articulating those differences visually- not an engineering problem.
this is what the article is advocating for.