(This is a lightly-edited version of a review I posted to the IntFiction forums during 2022's IFComp. I also beta tested this game, and haven’t done a full replay, so caveat lector)
There are various origin points for what we’ve come to call IF – Adventure, most obviously, but you can also trace choice-based games back to the print Choose Your Own Adventure series and its own early-20th-Century antecedents, and Aaron Reed defensibly started his 50 Years of Text Games series with the initial, purely-text versions of Oregon Trail. There is an eccentric uncle in the attic nobody really likes to talk about, though – or rather, aunt, since I’m speaking of the chatbot ELIZA. Viewed now as little more than a parlor trick – though how could it have been anything else, given the hardware constraints at its 1960s inception? – AI tech is finally catching up to the possibility of having a computer that can engage in a dialogue with you, even if the Turing Test is in no danger of falling anytime soon. So it makes sense that authors are now attempting to re-cross the streams and make a chatbot into a game, rather than something for pre-teen boys to feed dirty jokes into.
Of the runs at this idea that I’ve seen, Thanatophobia seems the strongest. I’m not equipped to evaluate the back-end of what makes it feel reasonably responsive, but there are some design parameters that are cannily set up to paper over the inevitable infelicities that will come up when trying to speak English to a robot. For one thing, the interlocuter character is set up as someone disoriented and not in their right mind, so the occasional odd interjection doesn’t seem too mimesis breaking. For another, the game’s built around a mystery with several pieces, so it’s less likely the player will spend so much time on one topic or area that they start trying increasingly-odd questions or statements. The author’s also done a good job of fleshing out various non-essential bits of backstory so that there’s room for the player to explore without quickly seeing the difference between the hand-tuned, critical path content and generic chatbot oatmeal.
The story being told here isn’t especially novel – there’s a little bit of a twist, but plumbing an allegory to discover someone’s hidden trauma is well-trod territory in IF by this point, albeit it does act as a clever homage to the psychoanalyst-aping roots of the chatbot conceit. And the characters inhabit well-worn archetypes without doing much to distinguish themselves. But for a formal experiment, keeping the narrative tame is probably the right call. Similarly, while the expected chatbot-friction is reduced, it’s definitely still there – but I do wonder how much of that would be smoothed if there were more uniform player expectations about how to interact with such things, much as there are by now for traditional parser games.
All told I found Thanataphobia a success, perhaps more intriguing for the directions it points to than for what it accomplishes in itself, but an entertaining way to spend an hour nonetheless.