Valve launches Steam Labs, showcasing "experiments" in improving discoverability

If you've ever been curious about what, exactly, it is that Valve does all day when nobody is looking, then prepare to be at least mildly sated. Valve has unveiled Steam Labs, a new website showcasing a range of experimental features, with a focus on discoverability, developed for potential use within its behemoth store. It's quite interesting!

"Every year," says Valve's introduction to Steam Labs, "we create dozens of experiments around discoverability, video, machine learning, and more. You know who we thought might enjoy seeing them? Everyone." The idea, it says, is to enable users to "try, share, and break" these experimental features, then give feedback directly to the developers that created them.

"We're always trying new things with Steam," Valve continues, "but often only share them with the world when they're ready to be made a part of the platform. Steam Labs allows us to share these ideas earlier and improve them with your feedback before making them official."

Three "active experiments" are currently available for perusal, and Valve notes that some "may turn out great. Others, we may toss out. We hope that most will be improved with your feedback and go on to be a part of Steam."

First on the list is an experiment known as Micro Trailers, which, as its name implies, is able to automatically turn full-length trailers posted to a game's Steam page into brief, six-second video snippets. At present, the Steam Labs implementation of Micro Trailers sees them triggered while running a mouse across a page of game thumbnails, linked to their store page. It's actually a fairly effective way of getting some early, useful first impressions of a game without having to hop through to a store page and work through the usual array of assets.

And it's an idea that's expanded considerably in the second experiment Valve has on display, Automatic Show. This, as you might well surmise, is a half-hour machine-generated video round-up of game releases and game-related happenings.

At present, there's not much more to it than a series of quick trailer clips (although Valve does also include a prototype with commentary), but there's certainly potential in the idea of offering a more visual, auto-generated round-up of the latest game releases, and it's clearly a more effective way for customers to quickly digest what's on offer than simply working their way one-by-one through a text list of unfamiliar titles. Half an hour is probably overkill though.

Lastly, there's the Interactive Recommender, which, even in its early state, is pretty great. Using machine learning, the recommender generates a personalised list of games based on your play time of other titles. You're then presented with a number of sliders and filters, enabling you to weight results by popularity (from very popular to niche) or release date. You can also choose to only display certain tags or omit tags altogether.

For me at least, it's the popularity filter that makes this most interesting. As someone who spends a considerable amount of time browsing Steam, I'm pretty used to seeing the same games recommended over and over again based on my play habits. And it's refreshing to see a whole new fist of titles by way of the "niche" option, supposedly tailored to my tastes, that have previously slipped under my (or Steam's) radar.

Valve has shared a little more behind-the-scenes information about Interactive Recommender on its official blog, so it seems reasonable to assume that it might be preparing to do the same for all Steam Labs experiments. It's certainly worth having a poke around if you're at all interested in this sort of thing, and hopefully these experiments may, in the future, go some way toward resolving Steam's notorious discoverability issues.