It's been about a month since Steam introduced its new Interactive Recommender that uses machine learning to suggest new games based on the ones you've already played. There are plenty of feature requests being discussed in Steam's own discussions page for Steam Labs, but the Reddit community dedicated to Steam has started a thread that does the exact opposite.
One of the stickied posts in the Steam subreddit, posted four days ago, asks other Redditors to post games that they consider "hidden gems." Given that it's locked to the top of the page, it seems the moderators may plan to leave it running for a while to let the community call out some of its favorites. Getting crowd-sourced recommendations is no novelty, though it's interesting that users are making a concerted effort to push a human-curated list while Steam focuses on auto-generated lists.
Steam's Interactive Recommender makes use of a lot of data to tailor game suggestions to you based on games you own, but there's something to be said for a bit of human intuition and spontaneity. So far in my experience, the Interactive Recommender has been great at showing me games that I already own on another store or platform, which is the subject of one current feature request.
In contrast, there are a handful games I haven't heard of in the Reddit thread. Some are not obscure at all, like Spelunky, which was PC Gamer's 2013 GOTY. But then there's Anxiety which has only four reviews (three of them negative). Is it a hidden gem? I have my doubts, though the user who posted it says "the game can be really frustrating but it is entertaining and the mystery is a nice touch."
Near the bottom of the page is a game called Pid, a game by Might and Delight, better known for its Shelter series of games. Despite being familiar with those games, I'd never heard of Pid, Might and Delight's earliest Steam release from 2012. The Redditor who posted it makes no mention at all of the studio's otherwise well-known pedigree.
Recommendations from strangers may not be as mathematically spot-on as Steam's Interactive Recommender tool, but the thread is a fun reminder that machine learning can only do so much. Sometimes, rather than seeing games I'm statistically likely to enjoy, I want to be convinced to try something I wouldn't normally play.
The thread only has 33 comments at the time of writing, but hopefully it grows. If you're in the mood to share, let us know what your favorite hidden gem on Steam is.