top of page
Search
Writer's pictureelizabethbenker

Providing Transparency in AI-enabled Recommendations


This recommendation appeared in my LinkedIn Notifications feed today. And in my case, the suggestion is spot-on. I've been a huge LeVar Burton fan ever since his Reading Rainbow days.

Even though the content resonates, I was turned off by the wording. The phrase, "People like you..." gave me pause. What do you mean, "people like me?" Not only does this sound like the start of an insult, it also raises questions about how LinkedIn's AI-enabled suggestion algorithms are making decisions under the hood.

This wording violates a basic user experience principle for AI-enabled content: be transparent about why you're making the recommendation. Outside of the aviation industry, nobody likes a black box. It's essential to provide users with information about where suggestions are coming from to give people more confidence in the quality of the recommendation. This kind of transparency is also necessary to identify and root out bias.

I can think of several ways to word this suggestion that don't make me feel insulted and provide more clarity about why the recommendation surfaced in the first place:

How about:

  • People like you who support literacy organizations are following...

  • Other people who went to Northwestern University are following...

  • Your connections NAME, NAME and NAME are following...

  • People in your Star Trek fan group are following...

Any of these options would be more informative and make me feel more confident that the recommendation isn't perpetuating inappropriate stereotypes.

What wording resonates most with you?


Tags:

0 comments

Recent Posts

See All

Comments


bottom of page