I was assigned with the task of writing about a gaming platform for my Games Anatomy class. My jumping off point for this topic was listening to NTS radio, hearing the tagline “music selected by people, not algorithms”. NTS host shows where music selectors, artists and tastemakers put together playlists and mixes. This way you can find some obscure gems you may not have ever come across before. I know I’ve personally bought a record from an Italian contemporary jazz band who have 884 monthly listeners on Spotify because I heard a track of theirs on a Ballamii mix.
It brought me back to our class where we discussed how the Steam algorithm works. I started to conduct an experiment off the back of this. When I go on Steam and look at the games it recommends in “More Like This”, it’s hardly ever useful. See the games recommended under If Found, a 2D narrative game in a graphic novel style. The gameplay is limited to erasing and filling in diary pages. Three AAA games are recommended (Fig.1). Maybe Detroit become Human could be comparable since it’s very narrative driven with limited mechanics but I would never put these games in the same category. Unpacking works as a suggestion, I’m fine with this.
Fig. 1
Let’s take another game, Firewatch(Fig.2). Mouthwashing, it’s an indie game I guess I could see how you would get there. Cyberpunk, Skyrim and Spiderman? Similar to the low poly walking simulator where you play as a park ranger processing your wife’s early onset dementia? I’m not seeing the comparison.
Fig.2
Let’s try a curveball, Bury Me My Love. A narrative game where you can experience the story of someone fleeing war and what their texts to their loved ones might look like. It’s apparently similar to….Skyrim(Fig.3)? Walking Dead I guess I get it, it’s a branching narrative but still a stretch and somehow I doubt it’s similar to Slime Rancher, a game where you’re, yes you guessed it, wrangling slimes.
Fig. 3
Another game on my wishlist is Arctic Eggs. What are Steam’s suggestions for games like this? Oh of course, Cyberpunk again(Fig.4). I can see maybe how it could be similar to something in the broader cyberpunk genre but what human being would say, “Hey you liked Arctic Eggs? Try Cyberpunk!”
Fig. 4
Okay, okay let’s just do one more. Indie darling What Remains of Edith Finch?. If I were to suggest some games based on someone liking this I’d think about Gone Home, The Vanishing of Ethan Carter, They’ve All Gone to the Rapture etc. What do we have? Drumroll please…(Fig. 5)
Fig. 5
I did this for a few other games and some were better than others, what I did notice was that both Cyberpunk and Skyrim came up very often. When I open my front page, what do I see? Cyberpunk and Skyrim(Fig.6 & 7). How can a small creator compete when we can’t even get recommended alongside other offbeat, indie games? And apart from that these just aren’t very accurate or intelligent suggestions.
Fig.6
Fig. 7
This little exercise reminded me of when, over the summer, I was desperate to get out of my barista job and I wanted to work from home so I applied for a remote job training AI. One of the exercises asked something along the lines of, “Imagine you had a friend who just finished reading Phil Ivey’s biography and really enjoyed it. Would “The Wager” by David Grann be a good recommendation for your friend?”. Now, for context, Phil Ivey is a professional poker player, The Wager is a non-fiction essay about a mutiny aboard a ship called The HMS Wager. Let’s see what Chat GPT thinks…
So even after explaining The Wager is about mutiny on a ship, Chat GPT still thinks this is a relevant suggestion? If someone told you they liked Michelle Obama’s autobiography, Becoming, you would look insane to suggest “If you liked that you should check out Metamorphosis by Franz Kafka because it’s also about transformation and adapting to change”. This brings me back to the NTS tagline “music selected by people not algorithms”. There’s a reason these places are hiring humans to improve LLMs. A human could tell that recommending Slime Rancher underneath Bury Me My Love is bizarro.
References:
Carless, S. (2024) Revealed: New Data on Steam ‘wishlist conversions’ & the hit-driven market. Available at: https://newsletter.gamediscover.co/p/revealed-new-data-on-steam-wishlist (Accessed: 22 January 2025).
Grattarola, M. (2024) Games Anatomy - Week 8, Powerful platforms or platforms of power.
zukalous (2021) How steam works against small games, How To Market A Game. Available at: https://howtomarketagame.com/2021/08/09/steam-hates-small-games/