SJT Text Adventure Game

Inspiration

We were tired of boring situational judgment tests in internship applications and thought they needed a revamp to be more enjoyable in the internship application process. We also took inspiration from old text-based adventure games, which have recently been repopularised by AI Dungeon.

What it does

The code puts you into two random stories and gives you choices, you choose each option and follow through the story, at the end you are given scores on various characteristic traits and an AI generated feedback on those scores based on your responses. We also attempted to create an algorithm that predicts how you will respond based on your previous responses and historic records of people playing the game, which can be left to better train over time.

How we built it

We wrote it using basic python modules like tkinter and we decided instead of making our stories AI-generated, we would take the creative liberty to write them ourselves. We then added in a couple AI models to give feedback and several other interactive features to make it more immersive like the old text based games felt. We felt that even though our knowledge of coding was limited, the simplicity made it seem more authentic to the old retro text games.

Challenges we ran into

We are all physics students so our knowledge of coding is not as great as a lot of other participants in durhack this year. As a result, many errors gave us many headaches. Also taking on the responsibility to write our own stories was slightly more time-consuming then we initially gave it credit.

Accomplishments that we're proud of

Our story about the space mice we are particularly proud of because it is creative and original and we like how the two stories loop into each other if you choose the right passage.

What we learned

We learnt a lot of skills, it was many of our first times integrating AI models into our code and was some of the most complex code many of us have produced since beginning to learn python both before and during our degree. As a result, we took away many valuable skills from pursuing things through error messages.

What's next for SJT Text Adventure Game

If the machine learning program works, then hopefully we can ask people to play our game and watch as it produces branches of probable paths based on previous responses, and thus we can perhaps learn a little bit about how people might make similar choices depending on their characteristic traits.