Procedurally Generated Points And Clicks: Symon

First-person lying still.

I was recently pointed in the direction of Symon, a free experimental point and click adventure from Singapore-MIT’s game laboratory, Gambit. The idea behind it is to see if it’s possible to create a procedurally generated narrative adventure, with unique puzzles. Which is quite an ask. The results, they’re an interesting combination of cheats and potential.

You play as a man paralysed in hospital, presumably dying. And from this perspective, you are naturally in fairly sad territory. Each adventure you generate is a dream, and as such embraces dream logic. And there’s where the clever cheats come in. Each time you play, and I’ve played it through six times now, it not only combines together a different collection of scenes, but also a different selection of puzzles using a small set of objects. There are also three main scenarios it picks from, mixing things up further, although for each you’re gathering three objects that represent memories from your past, gathered for a version of yourself in a hospital bed within the dream. Following me so far? Good.

Each location has two impossibly standing doors, linking them altogether, a final door not unlocking until all three objects have been collected. And to get them, it’s a fairly elementary inventory puzzle situation. Some girls with a balloon want a sad storybook, but the one you’ve found is cheerful. So you apply dream/adventure logic and click it on some bitter water in a sink. The combination creates a morose storybook, and the girls reward you either with another item used in a similar way elsewhere, or one of the three memories.

Next time you play, they may want a happy story, or some bitter chocolates, or they may not be there at all. Then repeat that idea for five or six locations, and you’ve got something approaching a unique situation each time you play, albeit with extremely similar motifs each time.

The result is mixed. The first time through it’s an interesting experience, a very short, novel adventure, where dream logic makes for an interesting twist. The second time you’re impressed at how different it is. And the third all you can see is the matrix. There also appear to be some issues with its letting you go down blind alleys, perhaps turning some roses blue before realising you need them red, and having no way back other than restarting. Restarting five minutes’ play admittedly, but it’s still a bit crap to do that. But what it does reveal is a great deal of potential. The internal ‘logic’ is consistent, and if the idea were taken much further, with far more scenes and a vast number more variables, you could end up with something that’s genuinely generating a playable experience that would be utterly unique each time. However, as I mentioned before, dream logic allows quite a lot of cheating in design. It’d be interesting to see if it were possible to create something that didn’t rely on such obscure combinations.

As a prototype, Symon is really exciting. And as a game itself, it’s a surprisingly emotive vignette those first couple of times through. It plays neatly in a browser, although while the music is absolutely lovely, there’s no flipping way to shut it up when you’re in another tab.


  1. Mike says:

    As you say, over time it becomes possible to ‘see the Matrix’ underneath it all, and I think proper procedural content for point-and-clickers would need to be at a richer level of puzzle design – rather than this-on-this-is-this. But that’s what this research is ultimately headed towards, I guess.

    There’s another game – Stranded In Singapore – which doesn’t have the strength of design, but presents the procedural technique in a less cheat-y scenario (no dream logic this time). You can definitely see the idea evolving.

  2. cmc5788 says:

    The problem with designing something like this on a scale larger than that of a prototype, is that procedural interactions tend to become exponential. So either you restrict that and end up with a very rigid experience, thus defeating the point of a procedural adventure to begin with, or you end up with something like… oh, let’s say, Dwarf Fortress.

    • Mike says:

      That’s the point, though – interactions become unwieldy, but artificial intelligence is there to reel it in and keep it sane. It’s why I think this research is really important.

  3. pilouuuu says:

    This seems much more interesting than anything that happened in adventure games in the last… 25 years? Alongside with Facade this shows that it’s possible to create interactive non-linear adventure games with true replayability. All games could benefit from this in a time when story driven games seem to be more linear and non interactive all the time.

  4. celozzip says:

    hasn’t this been done before? what was that black and white adventure game released a while back? it was different every time i’m sure of it… sorry for the lack of any real info in this post, maybe someone else can think of it.

    • celozzip says:

      nevermind i just played this and realised this is the one i already played, it’s from 2010. no edit button on this site?

      • TechnicalBen says:

        Yes there is. No idea why it is not showing for you. :(

  5. Harbour Master says:

    Such strange timing. I have an interview up with Symon product manager Clara Fern├índez-Vara tomorrow and thought I’d be introducing a bunch of people to Symon for the first time. But nooooo RPS has to jump in =)

  6. Acorino says:

    It’s a bit crap that you can get stuck in dead ends, but otherwise this is quite something! It works really well, actually.

  7. Premium User Badge

    Hodge says:

    I love seeing people do this kind of thing. It never quite works properly, but I’m always amazed that it works at all. I have a great hope that procedural design (if that’s what we can call it) will become massively transformative once it hits maturity, kind of like 3D did in the early nineties.

    @Harbor Master Thanks for the heads up about the interview!

    • The Random One says:

      I also love procedural design, but I hope it doesn’t do what 3D did in the early nineties, that is, to create myriad of new genres at the cost of destroying so many of what was already estabilished, to the point that it took over ten years for 2D design to be accepted again.

  8. TechnicalBen says:

    I’m much more interested in narrative really. But, this game does leave the scope for the player to add their own narrative (like they do in Minecraft).

    The interesting thing is, like mentioned above, games like Dwarf Fortress Adventure already do this. The narrative and story and game logic in that one is amazing. There might be a mission that involves seeking revenge on a group of outcasts who stole from one town, killed a certain leader, then upset a wizard. For a game like Fable that all sounds normal. But when you realise those things play out in the game and are dynamically generated, you realise the depth Dwarf Fortress goes to.

    It’s just a pity no one seems to translate such depth to any other game. :(

  9. Apples says:

    I think projects like these are attacking the problem from the wrong perspective. Let’s be honest, the interesting parts of this were hardcoded by humans – the effects of the various items, the item descriptions, the dialogue, the graphics, the overall story… the only thing this seems to do is randomly choose some hardcoded end goals and then randomly place hardcoded objects that have the right hardcoded effects when combined. It’s a decent technical achievement but I don’t see where this type of project can go, apart from increasing the amount of random events and interaction rules, or being used as part of a mostly human-made game; we don’t have anything like the sort of AI that would be necessary to create a properly unique, complex and sense-making story, and developing that kind of AI will not be furthered by projects that just define lots of rules. Right now the best we can get, like this game, is about as replayable and unique as an FPS with random enemy spawning – it is technically different each time but it’s always the same because it’s always operating within the same set of rules and with the same set of resources. DF has many, many, many more rules than this game and it still cannot tell a story well, and certainly not on the individual human level that adventure games are usually operating on – no matter how randomly and logically events proceed due to all the defined rules, a human storyteller and a lot of invention and extrapolation is required to make DF stories enjoyable to read. The best procedural narrative games now are ones that give ‘building blocks’ of events/situations and let the players construct a real narrative – the game does not tell the narrative, humans still do all the real work.

    I’m sure it’s doing some clever sort of back-chaining things to generate the sequence of item uses, but I just don’t think rule-based AI is ever going to be much use for narratives. Maybe I am being too negative but I can’t see it happening.

    edit: damn that’s a big paragraph. I feel like Wulf 2.0.