The Test

GAME LINK – PASSWORD: cs377g

 

Game Overview

At its highest level, my game is, as the name suggests, about a Turing Test. You are a participant in a study conducted by Parallax AI, a mysterious and wildly successful tech startup that pushes the frontiers of artificial intelligence. Your task of distinguishing the AI from the human in a chat conversation seems simple enough, but doubts about the nature of the study and of your conversation partners begin to make the situation less clear. However, the company still expects an answer to the question of which is artificial intelligence, and your choice could potentially have serious consequences for the other two entities.

 

In the beginning of this sprint, I had a lot of trouble actually settling on a story idea I felt comfortable in pursuing. My original premise was fairly different, focusing instead on the effects of climate change and the recent redirection of conversation in some circles from saving our planet to discovering alternative ways to keep the human race alive, such as by populating a different planet. However, I wasn’t sold on my premise and my ability to tell a simple, cohesive story within the time allotted. I decided to pivot, instead turning to the idea of a punchier, more character-driven story in The Test. I’m very fascinated by both the capacity for AI to develop personhood in fiction as well as by unreliable storytelling, so I was excited to use both in my plot.

 

Version History

My indecisiveness continued throughout the rest of my work on this project. As I switched premises during our first, tiny playtest in class, the main feedback I gathered was to be very clear on the stakes early on for the user. Without this, it was suggested, players may not have the same emotional buy-in to care about the characters they interact with, and as it is a game told purely through dialogue between the three central characters, a lot of the experience would be lost. I took this advice to heart, though I struggled throughout the entire process to implement it, as I will discuss later in the reflection.

 

My choice of game engine also flipped throughout development. I initially tried to implement it in Inform 7 as a parser fiction. I was really drawn by the interaction form factor, in which players type natural language to further the narrative. I thought this was a natural fit for my game, as I initially planned for most of the game to be through typed conversation, with a break to explore the physical space of the Parallax AI headquarters. However, using Inform 7 posed some challenges in representing the gameplay as I imagined it, namely through inputting natural dialogue and receiving natural replies. This is not how Inform is intended to be used, as its main strengths are in its ability to handle spatial navigation, inventory management, and flexible command input. I, however, planned to use none of these features and instead focus on the less well-supported aspect of dialogue. As this textual input was the only real part of Inform that was suitable for my game, I decided it would be much more worth my effort to instead develop in Twine with branching, fixed dialogue options to start, with the opportunity for change to textual input as a stretch goal. 

Along with this change, I decided to narrow the scope of the game even further to just the chat dialogue between the study participants, removing the aspect of physically exploring the company building. This was partially due to player feedback to focus more in on the main story aspects and deliver additional lore about the company through that same communication channel and partially due to the timing constraints of the game’s development. It was also useful to the narrative, as described below.

 

After settling on Twine, I focused more on developing and iterating on the narrative. Each of the following playtests was done by one undergraduate student at a time. It was suggested in my teaching team feedback to explore the player’s own sense of humanity through the lens of the threat of artificial intelligence, something I was also eager to include in the game. While my initial concept for the game placed the main character firmly as human, I decided to allow for more ambiguity to heighten the uncertainty of the player, further blurring the lines between human and AI. The constraint of the game to just the computer screen interface rather than the physical world as well helped add to this uncertainty about the main character’s truth. When the player can’t move from the computer screen, it begs the question of whether this is a gameplay choice or a narrative clue. I also added seeds for player doubt within the story itself through the unreliable narration. Playing the game, you may have noticed at certain times the participants in the study, Ronnie and Fritz, suggest that the main character could be artificial intelligence rather than them. There also moments of contradiction that can be found by replaying the game in the stories presented by Ronnie and Fritz. One such example is in the description of their earliest memories — depending on if they know a Turing Test is taking place or not, they have different responses. I wanted a small, visible contradiction to represent the lack of trust the player can put in the stories they’re being fed from any of the three sources, and the overall ambiguity of the situation. 

Players seemed to enjoy this ambiguity and, surprisingly, liked not knowing the answer definitely, even by the end of the story. This was not expected by me, and I was worried about providing a satisfying explanation of the mystery by the conclusion. A problem I often run into when trying to tell a mystery, even one as small as this, is that it can be difficult to top whatever theories your players can come up with without disappointing them. To this end, I decided to keep the ending ambiguous rather than create a definitive right or wrong answer. While there are three endings, none of them confirm for certain whether a certain character is an artificial intelligence, no matter who you pick. I believe this choice aligns more closely with the dystopia of unavoidable, ever-present AI that I wanted to display more than any other I had considered prior to this.

 

From testing, it is less clear whether players felt invested in Ronnie and Fritz’s characters or motivations. As I mention below, I struggled with making them likable enough characters with different voices and personalities, and it is possible this aspect fell flatter than intended. I believe with more dialogue available for each character, they could feel more lived in. This was suggested to me during a playtest with a friend, and I agree with it entirely.

Conclusion

There’s much I’d do differently if I were to continue working on the game, most of which is additional narrative and presentational polish. At the start of development, I had much grander visions for the scope of the game. Given more time, I’d want to add more questions, dialogue branches with unique responses based on previous conversational topics already covered, and more layers to the mystery and lore of the world. For example, a brief break in conversation to explore the contents of the computer you are communicating from, gleaning clues in a mixed media format, could be a great bit of variety. These clues could, in turn, unlock hidden conversational topics to discuss with the other characters, additionally adding more replayability. I’d also love the chance to add additional bells and whistles like animated typing and screen effects, as well as tighten up and revise the conversational passages to be more true to character and narrative.

Overall, however, I felt I learned a great deal about how interactive fiction differs from more linear, novel-style writing. I was excited by the many possibilities for replayability and creating different experiences based on playstyle. I’d love to dive into all of this even deeper if I were to start again.