what-could-go-wrong-ai

What Could Go Wrong - AI Edition

This repository hosts the files used for the What Could Go Wrong? digital card game. Nikolas Martelaro and Wendy Ju developed and presented the game at a Workshop at AutoUI 2020. The game was then modified by Nik and his team at CMU to be more generalized for AI-based systems.

The digital card game can be loaded into and played on PlayingCards.io. Custom cards can be added using the csv files for prompts and responses

Abstract

While AI has the potential to greatly improve our daily lives, there are also challenges and potential downsides to these systems. This game aims to foster discussions about the potential negative aspects of AI in hopes of surfacing challenges that should be considered during the design process rather than after deployment.

Game Setup Instructions For playingcards.io

  1. Download this repository
  2. Go to https://playingcards.io/game/standard-deck and choose “Start a Blank Room”
  3. Enter the virtual card table
  4. Click the Edit Table icon in the green toolbar
  5. Select Room Options
  6. Select Import From File
  7. Upload what-could-go-worng-av.pcio
  8. Click Edit Table to exit editing mode and go into gameplay mode

Gameplay Instructions

Setup

  1. Each player draws 5 white response cards from their stack
  2. Click the spinner to randomly choose the first Card Czar

Each Round

  1. The Card Czar draws a black prompt card and reads it aloud to the group
  2. All other players select 1 white response card from their hand and place it face down in their designated slot
  3. The Card Czar flips and reads each white response card out loud
  4. The Card Czar chooses the response card they think best answers the prompt
  5. +1 point goes to the player whose card was chosen
  6. Group Discussion: The group discusses what else could go wrong based on the chosen card
    • Players can award +1 point to anyone who makes a good point during discussion
    • Discussion should last a few minutes
  7. Next Round: The next player becomes the Card Czar
    • Click the “Deal” button to give each player a new white card
    • All players should have 5 cards in their hand at the start of each round

Game Flow

Important Notes

Video Demonstration

What could go wrong card game demonstration

Adding new cards

Edit the prompts.csv and responses.csv to add new cards to the decks. Follow instrcutions for adding new cards here: https://playingcards.io/docs/custom-decks

Creating Printable Cards

If you want to create physical, printable cards from the CSV data, you can use the included generate_cards.py script.

How it works

The script reads the prompts-ai.csv and responses-ai.csv files and generates high-resolution PNG images for each card:

Requirements

Usage

  1. Ensure your CSV files are in the repository root
  2. Run: python generate_cards.py
  3. Generated cards will be saved in the CARD IMAGES/ folder:
    • PROMPT/ - Black prompt cards
    • RESPONSE/ - White response cards
    • Card backs and instruction card

Output Specifications

The generated images can be sent to a print service such as The Game Crafter for professional card printing.

Suggested Citation

Nikolas Martelaro and Wendy Ju. 2020. What Could Go Wrong? Exploring the Downsides of Autonomous Vehicles. In 12th International Conference on Automotive User Interfaces and Interactive Vehicular Applications (AutomotiveUI ‘20). Association for Computing Machinery, New York, NY, USA, 99–101. DOI:https://doi.org/10.1145/3409251.3411734

Bibtex

author = {Martelaro, Nikolas and Ju, Wendy},  
title = {What Could Go Wrong? Exploring the Downsides of Autonomous Vehicles},  
year = {2020},  
isbn = {9781450380669},  
publisher = {Association for Computing Machinery},  
address = {New York, NY, USA},  
url = {https://doi.org/10.1145/3409251.3411734},  
doi = {10.1145/3409251.3411734},  
abstract = { While autonomous vehicles have the potential to greatly improve our daily lives, there are also challenges and potential downsides to these systems. In this workshop, we intend to foster discussions about the potential negative aspects of autonomous cars in hopes of surfacing challenges that should be considered during the design process rather than after deployment. We will spur these conversations through a review of participant position statements and through group discussion facilitated by a card game called “What Could Go Wrong?” Our goal is to consider the autonomous vehicle’s benefits—improving safety, increasing mobility, reducing emissions—against potential drawbacks. By identifying potential harms and downsides, the workshop attendees, and the AutoUI community more broadly can design well-considered solutions.},  
booktitle = {12th International Conference on Automotive User Interfaces and Interactive Vehicular Applications},  
pages = {99–101},  
numpages = {3},  
keywords = {game with a purpose, failure modes, autonomous vehicles},  
location = {Virtual Event, DC, USA},  
series = {AutomotiveUI '20}. 
}

For some motivation on why we want to develop new hazard analysis games.

  title={Exploring Opportunities in Usable Hazard Analysis Processes for AI Engineering},
  author={Martelaro, Nikolas and Smith, Carol J and Zilovic, Tamara},
  journal={arXiv preprint arXiv:2203.15628},
  year={2022}
}
This site is open source. Improve this page.