On NLP and Games as
Philosophy
had some thoughts today, but I need to share some background first.
So the idea with Quell is to make a philosophical game. I was inspired by Heidegger’s conception of human emotion as revealing metaphysical truths to us, that the negation of logic needs a deeper grounding. “The idea of ‘logic’ itself disintegrates in the turbulence of a more original questioning. … [N]egation cannot claim to be either the sole of the leading nihilative behavior in which [our being] remains shaken by the nihilation of the nothing. Unyielding antagnonism and stinging rebuke have a more abysmal source than the measured negation of thought. Galling failure and merciless prohibition requires some deeper answer. Bitter privation is more burdensome.” (Heidegger, tr. Krell, “What is Metaphysics?”)
That was about a decade ago. I’ve been reading and writing and learning to code since then. This is my third, and most sustained, attempt to make this game a reality.
The following thoughts are about whether or not a ‘philosophical’ game is feasible at all. I’ve lately been experimenting with the idea of having a player interact directly with NPCs via a free text input. This has lead me down the road of using natural language processing (NLP) and research techniques in argument mining to interpret dialogue. But so far I’m not getting very good results, and I’m beginning to question if this is even the right approach.
—
NLP might not be needed. Removing it requires player cooperation – but all games do. I just finished playing Soverign Tower, which is an information-lock gmae. There are guides. Outwer Wilds has been solved for a long time. Perhaps the act of original philosophizing is a bit out of reach for now. But no literature has ever been so flexibile as to allow the player’s opinions to influence it. There’s always a thesis. Soverign Tower has Epicrates convicne hte player to tear down the tower. Outer Wilds discusses the end of a civilization and what comes next. Disco Elysium is a static story, where the politics are explored with no ‘true’ ending – a thesis to itself, a form of political nihilism? A study of the state? The insertion of the player’s opinion (as contra or tangential or irrelevant to the main thesis) destroys the work as a static object. So it ceases to be what it is. – Perhaps games as philosophy – in the sense of the stage upon which a player does philosophy, as a new, unique synthetic stance (and not in relation to an existing thesis) is impossible by virtue of the game as a static object. Which is not to say an artifact cannot be philosophical in the sense of proposing and defending a thesis. This is, after all, just what philosophy texts, as literature, do.
What distinguishes philosophy from science? – I raise this because a game which relies on free text would need to lay the tracks as it runs – the world would need to shift accordingly. NPC behavior &c would change, thus changing the way the story progresses. We give it a starting point and a history and it moves from there. (Earlier drafts of the game relied on this procedural nature, a la Dwarf Fortress or RimWorld.) But analysis of this is analysis of one, or a series of, possible worlds. Does philosophy proper need to be universal? Ideal, Platonic? Or immanent, Deleuzian? What would it mean for it to be immanentist? – to engage with the infinite rather than the ideal?1 The scope of logic, as a discipline, is infinite – all thought. Or, ideal thought. Where do the ideal and the inifinte distinguish themselves? Ideal as transcendent, above reality; infinite as immanent, within and composite of reality.
*
The way the static version of Quell would work is by presenting artifacts to NPCs. You get to know them and can contact them via your soul stone throughout the game. ITems may have strange consequences depending on context and who you interact with; human draam stuff. ARtefacts are tagged with concepts. No player notes describing the artifact and interpreting them; just my interpretations; the logic is there and static, the ideas are “objective.” Feels like a bit of a consession but this is at least a good first draft, and can be automated with NLP later if desired – if its not fun, or doesn’t meet the expected outcome (real engagement). – This requires some faith in the reader to engage genuinely – but all literature does. And it’s certainly less risky than free text input.
Mechanically, texts are tagged with concepts and stances about those concepts. Presenting them to NPCs yields new paths, like free text but far more constrained.
–
If you’re interested in Quell and it’s development, check out the wiki and the github org.