This week has the CMU Carnival. We had days off on Thursday and Friday. So we only have 3 days for development. This week, we finished a video for our website which includes our first three prototypes and playtest results. Also, for the reinforcement learning prototype, we finished the visualization and designed some levels to show different concepts of Reinforcement learning. For the VR painting prototype, we designed the user interface.
Final Video
We finished our first version of the final video, which includes our hypothesis, the first three prototypes’ demo video, the playtest result, and the conclusion. We will update the video when we finish the reinforcement learning prototype and VR painting prototype. Also, we will add our final conclusion when we finish the playtest for these two prototypes.
Reinforcement Learning Prototype
First, we visualized the AI agent, goal, punishment and Q value because it will make it easier for the user to understand the relationship between each element in the scene. We use dinosaurs to visualize the AI agent. For the goal, we choose meat. The punishment is the rock and lava. For the q value, we use the color of the grass to show it. It is common sense that the dinosaur wants to eat the meat and dodge the lava. So we think the user will quickly get it.
| AI Agent | Reward | Punishment | Q-Value |
 |  |   |  |
Also, we designed some levels to show different concepts of reinforcement learning. We also try very simple storytelling. So this time, our specific hypothesis is: Providing a relatable experience through narrative storytelling can help users to be inclined to learn about data/ML. Currently, we have 5 levels + 1 sandbox mode. Each level focuses on a different perspective of reinforcement learning.
| Level | Concepts of reinforcement learning | Story |
| 1 | Action | A baby dinosaur was just born and learned how to walk. |
| 2 | Goal | The dinosaur is hungry. The user needs to feed it. |
| 3 | Punishment | The dinosaur walked too close to a volcano. Users need to drag lava to the environment. |
| 4 | Greedy Algorithm | Two ways to reach the meat, which way will the dinosaur choose? |
| 5 | Multiple Goal | So many meats around the dinosaur, which meat will the dinosaur choose? |
| Sandbox |
Currently, there are still many things that need to be refined. First, we will design the story more carefully to make it more interesting. Also, we want to add sound effects. We will also re-design the UI next week. If we still have time, we will design more levels. Finally, we plan to have a playtest next Friday.
Painting in VR Prototype
This week, we confirmed the user interface of the VR painting prototype. The UI mainly has three components. The first component is used to choose the image. The second component is for choosing how many slices the user wants to see. The third component is for customization, such as changing the rotating speed. We plan to implement the UI and interactions by next Tuesday and we will conduct the user test next Wednesday.
