OVERVIEW: AI or Nay-I? is a single player simulation game where players examine how AI systems work and evaluate how they will be regulated in their own fictional nation. It is a transformational game designed to demystify AI and help the public foster a more informed view towards artificial intelligence's potential impact on human life.
ROLE: Lead Designer, Producer
TEAM: Jue Wang (UX Design, Art), Shitong Shen (Design, Programming), Zhiguo Lai (Design, Programming), Lotus Li (Sound Design)
DURATION: 4 months, 2019
CLIENT: Pamela McCorduck
TOOLS: Unity, Photoshop, Microsoft Office Suite
As lead designer, I held the vision for the project. More specifically, I devised the gameplay mechanics and narrative design. I produced all the written content in the game as well as all design documentation. I also wrote the copy for all promotional material.
As producer, I supported the team in realizing our vision, communicating with the client, and keeping our scope. I also maintained the project website including periodic blog posts on our development process (link below).
I. Research and Mission Statement
The team was tasked to create an interactive experience inspired by the work of author Pamela McCorduck. Pamela McCorduck is the author or co-author of ten published books. Both her fiction and nonfiction deal with aspects of science. This project focuses on her writings on artificial intelligence (AI). Pamela’s work explores everything from metaphysics to the lives of the scientists creating the systems. McCorduck spent her early career pulling on the shirts of important individuals, trying to pull their focus to technological developments that will have a monumental impact on human life.
We began the design process by conducting research. We compiled a list of the various uses of AI. It didn’t take long to discover our first challenge: scope. AI is a vast ocean of potential subjects for an interactive experience. We also conducted interviews with people, aged 21-45, to get their thoughts on the subject. We found a few common threads in the interviews:
People have a vague idea of what AI is.
Generally there is a negative to neutral disposition towards AI
People are fearful of the unknown power of AI
With that in mind, the team developed an initial mission statement for the project:
Demystify AI and foster public conversation about AI’s potential impact.
Given AI’s widespread impact, the team decided to develop a short mobile game, about 10 minutes, with a target audience ages 13-40.
Research and Mind Maps
The game, titled AI or Nay-I? , was designed with The Transformational Framework by Sabrina Culyba. This framework is a tool to help designers realize their transformational goals in their game. The transformation we sought is threefold, after the game players will:
Have a better understanding of how AI works.
Have a better understanding of the moral complexities of AI.
Feel better equipped to make judgements with regards to AI.
Then, to cut down scope, we decided to each pick a use of AI from our compiled list. Each designer then took a couple days to design a paper prototype based off of that use.
III. Paper Prototypes Playtesting
The three subjects we decided to prototype were:
Life as an AI Researcher
Multiplayer Environmental AI Simulation
AI Ethics Investigator
We playtested our prototypes with our fellow students at the ETC. We had around 10 playtesters for each prototype. To measure our transformational goals, we surveyed our playtesters to gauge their opinion of AI both before and after. After reviewing, the survey results, we determined that the prototype #3, The AI Ethics Investigator, best fit our goals both in its gameplay and transformational ability. We then began to expand on this design throughout a series of long design meetings.
User Journey Maps
First Draft Experience Maps and Sketches
IV. Initial Design
After these lengthy meetings, I produced the first draft of a design document. At this point the game structure was:
An AI Ethics Committee has assembled a focus group to discuss applications of artificial intelligence in computer vision. The player is a member of the focus group. At this meeting the focus group will discuss three applications of AI, specifically computer vision: Face ID, Tumor Recognition, and Surveillance. The moderator of the discussion will introduce reports for each application.
In each report, the player plays through a series of simulations that explore how AI works in each application. After each simulation, the focus group members speculate how and to what degree each application will affect human life.
After all simulations and discussions, the focus group fills out a survey detailing their thoughts on how mysterious AI is to them, how strongly regulated AI should be, and how beneficial AI could be for humanity.
With this structure in mind we began developing the Face ID simulation, conducting internal playtests every couple days or so. The simulation consisted of two phases. In the first phase, the goal is to collect sufficient data of phone’s user to use as reference for unlocking. In the second phase, the goal is to correctly identify if a person is the same as the one in the reference photos taken in phase one.
The Tumor Recognition simulation was structured similarly. In the first phase, the goal is to sort images of lungs into malignant and benign references. In the second phase, the goal is to correctly identify as many malignant scans as possible in the designated time.
Lastly, the Surveillance simulation even more closely resembled the FaceID sim. In the first phase, the goal is to collect as much data, face photos in this case, as possible in a given time of a street corner. In the second phase, the goal is to find a matching photo of a tagged person using the reference photos taken in phase one.
V. Iteration: Round 1
In addition to our internal playtests, we also met with a few special guests who offered feedback on the project as it stood. John Sharp, a visiting scholar from Parsons, and Anthony Daniels, actor and ETC special faculty.
John and Anthony both playtested our current iteration of simulation 1. We noticed that both Anthony and John tried to use the gyroscope in the data collection phase of simulation 1. In addition, they both struggled to understand what their goal and point of view was in the experience. The sorting phase went a lot smoother. After each playtest, we discussed the project concept and design. John advised us to better connect the facial recognition gameplay to the exploration of AI in the project. Anthony agreed with this point. Anthony’s discussion was geared towards our target audience and why we are specifically targeting them. His questioning, though intense, helped us hone our focus moving forward.
After the playtest sessions, we updated the control mechanism of the data collection phase to use the gyroscope on the mobile devices. Players would now be presented with a “blank” face that resembles a wire frame of the character’s face. Then players will collect data, by tilting the phone to rotate the face. The feeling of data collection more resembles painting on a 3d model. We felt this iteration better communicates the process of 3d model creation that the AI in mobile devices do in real life while also engaging the player better.
To address the confusion with point of view, we hope an opening shutter at the beginning of data collection phase will better inform the viewer that they are inside the device in this phase. This can also be reinforced by short animations before the minigame in the opening scene of the game.
In addition to mechanical changes, we further iterated on the story of the game. Instead of playing as focus group member, players are now the Minister of Technology for a fictional nation. As Minister, players examine uses of computer vision AI, the minigames, then sign mandates that either ban or allow the technology to be used without regulation. Then players are presented with the ramifications of their decision, in the form of a social media feed that includes news headlines and posts from the public.The team feels that this raises the stakes of the game and highlights the moral complexities that advancements in AI bring.
Beyond the changes detailed above, we also purchased art assets to save production time.
VI. Iteration: Round 2
Satisfied with the status of the phone security simulation, the team dove right in to the development of public surveillance simulation. We quickly found out that the time we allotted for the development of the last two scenarios was not large enough and we decided to cut the tumor recognition simulation. As a result, our game’s focus is further narrowed down to focus on facial recognition systems.
With this change in place, I designed the layout for the public surveillance level with the assets we purchased. I also produced the narrative design and wrote all content for the game. A full flowchart of the game’s structure was made shortly after the script draft was completed. The flow chart can be seen below.
We then proceeded to stitch together the project into our first full build to be tested at the ETC Playtest Day
VII. Iteration: Round 3
We tested the build at the ETC Playtest day with a large group of playtesters. We had 11 playtesters ranging from age 16 to 50. After running through the experience, we had the playtesters fill out two surveys. The first gauging how successful AI or Nay-I was at transforming the players while the other survey comprised of questions from the Schell Games Guide to Playtesting.
We also tested with several industry experts over the next few weeks. Jeff Litchford, VP of Deck Nine Games, praised the user interface and interactions but found the perspective shifts from Minister to AI Simulations and back to be jarring. Jeff also offered a helpful suggestion to ask the players “Are you sure?” an extra time after signing the directives. Later in the week, Joe Olson, Art Director, and Jonathan Brodsky, Creative Engineer, from Magic Leap stopped by. They enjoyed the game as well and suggested that there may be opportunities to infuse subtle color theory tricks into the public response tweets: like cool colors for positive responses, and warm for negative.
The team also showed the game to a couple faculty members for feedback. Jesse Schell encouraged us to include a feedback screen after the minigames to compare the abilities of the player to those of a real AI system. In addition, Shirley Saldamarco also completed a demo. Shirley advised us to be clear about the purpose of AI or Nay-I? when we are showing it. The game isn’t trying to convince the players that AI is good or bad, but rather portray the complexity of the issue and foster thoughtful discussion in the players.
We incorporated all of these suggestions into the next build of the game to be shown at Soft Opening.
VIII. Iteration: Round 4
At the team presented AI or Nay-I? at the ETC’s soft opening. Then, after reviewing the faculty feedback, we set off to polish the game before final presentations.
The week began with a large push to incorporate as much as possible into our build of AI or Nay-I for soft opening. The faculty praised the smooth interactions and consistent art style throughout the experience. Also, our decision to cut the scope and focus on a more quality, albeit shorter, experience was wise. The last improvements to the game before finals included:
I balancing the biases in the public response tweets in an effort to more fairly examine the debate
I reworked the the tweets to feel more personal to the minister in order to increase emotional reactions
We splitting the minister’s decision into two parts: a press release and the official directive. Each would have their own round of responses. This would better explore the debate while allowing the player to change their mind, now with more information to base their decision on.
And with those final polishes, our design journey ended!
The player is the Minister of Technology for the fictional nation of Genovia, In this experience the Minister is tasked with examining two applications of AI, specifically facial recognition: Face ID and Public Surveillance.
In each report, the player plays through a series of simulations that explore how AI works in each application. After each simulation, the Minister first announces their decision in a press release, then they receive feedback from the citizens of Genovia. Then using what they learned from the first round of responses, the Minister signs a mandate to ban or allow the specific use of facial recognition technology. After the mandate is signed, the Minister examines another round of responses, this time in the form of news headlines, to reckon with the ramifications of their decision.
After all simulations and mandates, the player fills out a survey detailing their thoughts on how mysterious AI is to them, how strongly regulated AI should be, and how beneficial AI could be for humanity. The players also see how other minister’s have voted in their playthroughs.
A link to the full design document can be found below: