P2: Erosion

Game

My game can be accessed here. I would recommend playing the game prior to reading the write up.

Overview

In Erosion, players will take on the role of June Weng, a 24-year-old software engineer at Cerebra, a fast-growing startup that saves people’s memories and lets them re-experience them in VR.

June joined Cerebra after she graduated from college two years ago and is the prodigy of Cerebra’s founder, Robert Fuentes. The two of them have formed a close bond (akin to a father-daughter bond, although some playtesters sensed something else there haha), particularly because she has difficult relationships with her family, and he lost his wife and kids in a car accident years ago (which prompted him to start Cerebra).

Cerebra’s pitch is simple: make memory a place you can return to. From enabling users to sit with loved ones they’re grieving to aiding in dementia care to helping users process past experiences, Cerebra turns memory into a space for relief, learning, and feeling—whenever you need it.

At a party one night, June meets Elena, a Department of Defense (DoD) analyst who has found troubling information on how the DoD is using Cerebra technology. As the game progresses, June learns more about how the DoD wants to use this technology for mind control and is testing prototypes on teenage girls of color. June is pulled between a desire to do better for the world, not hurt it, and the job and father figure that she loves.

Erosion‘s objectives are to (1) explore the potential consequences of the rapid, unchecked progress of technology and (2) build an emotional connection between the issue and the player. As a part of the second objective, the game explores the United States’ history of testing medical advancements on vulnerable women of color: a theory it explores is that as technology becomes more integrated with the human body, institutions will turn to human testing once again, and those subjected to it will again be vulnerable women of color.

Starting with the third playtest, I ask each player how well they feel the game fulfills the two objectives I outlined, and their assessments demonstrated to me Erosion‘s improvement over time through playtesting and iteration.

Figure 1 maps out the game’s key choices and possible outcomes.

 

Figure 1 : Branching Choice Map

Testing and Version History

Note that recordings for Versions 1 and 2 are only of gameplay and not feedback.

Version 1

My first version didn’t actually tell the story from June’s perspective, but rather from the perspective of a girl who had gotten kidnapped to be part of the testing program. I had implemented one scene, where she’s working a shift at the local diner. It has three different endings: her getting kidnapped, her going home early, and her just continuing her work at the diner.

My playtester was a male CS 377G student studying AI at Stanford, giving him more background on tech, tech ethics, and game design than the “average” person. He got really into doing the voices for the different characters (starting ~1:00 in the video), signaling good engagement with the story!

When discussing feedback, he was interested in the potential other endings and went back to try them out. For Version 8, encouraged by a suggestion from Christina, I would ultimately make it so that players had the option to go back to the most consequential choice they made and see different endings.

For Version 1, however, my playtester observed that the endings for this scene were very different after just two choices, and it could be difficult to keep building off of these. In the next version, after getting some advice from Christina, I mapped out my story from start to end (see Figure 2 for some brainstorming), which resulted in writing a completely different story, to ensure that the branching I implemented would get me from the start to the various endings I wanted instead of getting lost along the way.

Figure 2: Brainstorming for Version 2

I also used stats (e.g. personality traits like compassion), but he found them confusing since they didn’t serve much of a purpose yet. Butch also mentioned that I should be intentional about how I implement stats since they can get confusing. In the next version, I realized they did not add much to player experience, so I eliminated them.

Version 2

Version 2’s premise is the one I’d stick with for the rest of the iteration process. I playtested it with a 22-year old female CS Coterm student, who has specifically taken tech ethics classes. This would also give her a solid background on the topic.

She played the game twice, doing nothing the first round and publishing an exposé on Cerebra the second time. She said she felt more decisive in the second round because she actually ended up doing something but not so much in her first round. My intention with the “do nothing” option was for the player to really weigh June’s personal incentives (including potential mistrust of Elena) versus an imperative to do the right thing, but it seems that did not come through, so I worked on building up the internal reflection and relationships with various characters for the next version. To facilitate this, I also tried adding back in stats to further build up a sense of who June is and how she values her relationships with specific characters.

I also observed as I was watching her play that there was a lot of reading without being broken up by choices early on, so I added some more choices to break up the text so that the character could feel like they were making decisions instead of just decisions being made for them.

I went back to the drawing board for the next version to ensure that my choices and endings felt meaningful (see Figure 3 for a snapshot of brainstorming).

Figure 3: Brainstorming for Version 3

I wanted to do analog so that I could visually see how everything fit together, but the sticky notes were getting messy, so I switched to digital (but still simpler format than ChoiceScript). As you can see in Figure 4, this also got a bit messy, but I could more visually see what was going on and include more meaningful choices.

Figure 4: Version 3 (Figma flowchart available here)

Version 3

I playtested Version 3 with a female CS377G student (CS Coterm, Symsys undergrad). Her academic background gave her a good understanding of technology as well as its intersections with humanities/social sciences fields (which is what the story explores). She does not like long-form fiction, which was the direction I was trending in (even with the flowchart), which did put her slightly out of my target audience.

Playtesting was chaotic because of the flowchart format but otherwise went smoothly. She struggled with one of the last few choices (whether to trust someone or not) because that character wasn’t characterized enough in the story, indicating that I needed to continue building up my characterization in the next version. She was also surprised by the ending she got since she didn’t realize it would be the result of her last choice, which indicated an issue with endings once again.

These pointed to perhaps a deeper issue with my game: I had been focused on mapping out branchings but not so much on building up the story itself to give those branches meaning. Butch suggested writing out the entire story with just one pivotal choice before adding in more choices, which is what I implemented for Version 4 (one branch has an additional choice on it, but it is for the most part just 1 choice).

My playtester did like how some things could only be figured out by doing another play with different choices, which couldn’t be implemented in Version 4 but was kept in mind for future versions. Originally, this was only referring to information surrounding the DoD program, but I would make the decision when building Version 6 to have this apply to other information as well.

She did not like the long paragraphs, so in Version 4, I would break them up into multiple pages, so the user has to click next and engage with the story throughout the text. There was one technical scene that she did not understand, so I did didn’t include explicit technical jargon in the next version.

Additionally, although she did not hit any pathways affected by stats, she liked the general idea. I’d end up taking them out for Version 5 due to later feedback, but this point of feedback made my decision harder.

She also liked the quick epilogue for each ending, although Butch suggested leaving the player wanting more. Initially, I tried to strike a compromise between the two by giving some information but not all, but I’d ultimately eliminate future information altogether in response to feedback on Version 7.

Butch also gave two pieces of feedback that were helpful. First he pointed out that he wasn’t so sure Elena would share her information so freely at a party with someone she just met, so I made that part more realistic by adding tension and making the player work a bit to get the information from Elena. Second, he suggested being consistent about whether an action is assumed completed because a user makes a choice versus the choice happens and then the action is actually depicted. I chose the latter, although there are some cases where it would be very clunky to do so, but I would playtest with these choices multiple times and no players said anything about them or exhibited any signs of confusion, which is why I was confident proceeding with the slightly less direct but cleaner story.

Finally, as outlined in the overview, my assessment asked players how well the Erosion accomplished its two objects: (1) exploring the potential consequences of the rapid, unchecked progress of technology and (2) build an emotional connection between the issue and the player. My playtester said that my game accomplished (1) but she didn’t necessarily feel the emotional weight I was hoping to impart because it felt too obvious. It would be more meaningful if I built up relationships and the importance of working at Cerebra to June, so I did that for the next version. Partly due to this feedback, I also decided to change the place names from fictional to real to keep it more grounded in the real world to hopefully build up that emotional aspect.

Version 4

I playtested Version 4 with another female CS 377G student of a similar demographic to my Version 3 playtester: familiar with tech/tech ethics and does not like long-form fiction.

The game played smoothly for the most part, but she got confused with a transition from a flashback to the present, so I made that transition more seamless and established setting better. Additionally, two of the characters flirt a bit when they first meet, and she was confused about that. It serves no purpose later on, so I just eliminated it in the next version SHe also didn’t understand a quick escalation in tone in a conversation, and I realized it was not as realistic as I had thought it would be, so I changed the tone of the conversation.

Key pieces of feedback (all implemented) are

• break up the text even more (i.e. paragraphs into smaller lines, long pages into multiple choices)
• characterize Robert more (so I added more scenes with him and fleshed out existing ones)
• make the testing being done on teenage girls more emotional (at 13:27)
• add more tension in June’s and Elena’s whistleblowing processes (at 16:00)
• include more parts where June is stressed at work, can’t eat, can’t sleep, is really paranoid, etc
• a lot of the technical jargon isn’t necessary: I should just keep enough to show that June is a technical person but not get super into it

Given that I had only used 1-2 pivotal choices in this version, I asked her where else she’d like to see choices, and she suggested (at 13:03) having June decide whether she wants to open an envelope she finds in her boss’s office (in Version 4, she does not) or not as well as decide whether to believe someone who is telling her that her company is involved in wrongdoing or not (in Version 4, she believes automatically). I implemented these choices, and to influence player choice’s less, I also got rid of my synopsis at the start where I hint that something bad might be going on so that the user decides for themselves whether or not something is, instead of being given it as a fact from me.

Another change I made was eliminating the VR Glasses people in this world use for communication and instead using a telepathic chip so that it’s less conspicuous. The purpose of either was simply to indicate that we are further in the future.

Since I was still going back and forth on stats at this point, I asked her how she’d feel about me adding them in, and she said she didn’t feel it was really necessary to meet my game goals and engage the player. I weighed the feedback from Version 3 and Version 4 playtests and ultimately decided not to include stats in Version 5 since I also personally felt that the changes they led to didn’t feel meaningful enough. I also addressed characterization and player agency in other ways that felt more aligned with what I was going for with the game (e.g. through detailed flashbacks, inconsequential choices that enabled player exploration, and consequential choices that directly mapped to, at this point, 9 different story outcomes).

Finally, for the assessment (at 17:23), she said the first objective was accomplished but she didn’t get a devastating ending and actually felt relief in the end. In the next version, I flesh out the emotional elements of the story and make the endings all more complicated (i.e. even the “good” ones involve some form of loss, even if smaller than the “bad” endings).

Version 5

I playtested Version 5 with a 22-year old female History Coterm student. She does a lot of reading for class due to her degree program, but she doesn’t like reading heavy games. She does, however, enjoy IFs and is familiar with them.

I noticed that there were a few moments where it was unclear what was going on. For example, since I got rid of the synopsis introducing players to the story, they no longer knew that June had been at Cerebra for 2 months, so when I used “In her third month at Cerebra…” to transition to a flashback, my playtester didn’t know it was a flashback (at 1:40). Getting rid of the synopsis also meant that she didn’t know what Cerebra actually did (at 3:34). As a result, I decided to add a synopsis back in for Version 6 since the only reason I took it out was that it hinted at unethical behavior at Cerebra, so I just inserted the parts back in that say who June/Cerebra are in neutral terms.

She was also confused at two points where June/Robert and June/Elena are discussing technology, so I completely took out technical terms and made explanations clearer. I noticed that she gave a great deal of thought to whether she should enter Robert’s office or not (at 19:08), which indicated to me that it was a good choice to include since it engaged the player. However, she did not know there was surveillance before she made that decision, something June would know, so I added that in.

When she got her final outcome, she was a little confused why Robert betrayed her until I explained (June was vague about something important, and Robert didn’t trust her anymore i.e. since she’s not telling him everything, he doesn’t trust her to work on something so sensitive with him anymore), so I made this clearer in Version 5 that being vague will be seen as a sign of being disloyal but will protect someone else, so the player has to decide which is better.

Ultimately, she found that it was a fun moral game to play (at 33:45). I also asked her how she felt about the amount of text on each page, given that that had been a problem in the past two playtests, but she said that there was never a moment where she thought it was too much .

However, while she understood that I needed to build up context before the player could make consequential choices, she recommended including some more choices earlier on, consequential or not, to get the user used to the interface. For example, she suggested giving the option for June to talk to more people at a party she goes to and use that as a way for the player to get to know her more, if they chose (see Figure 5). I added that in, as well as 2 other choices of the same style.

Figure 5: Choice to talk to more people at Nico’s party

These choices would add more of players-only-getting-information-through-certain-pathways, which my Version 3 playtester had explicitly mentioned liking. Additionally, even though I had decided not to include stats for this version, I was still wavering a bit on that decision, and this new approach would allow me to enrich the personal aspect of the story in another way (instead of stats), which ultimately solidified my decision.

She also said that when she was given a choice, since ChoiceScript automatically highlights the first choice, she felt that there was some bias towards that choice. In Version 6, I specified at the start that choices are not shown in any particular order so that users know there is no bias, but I would later eliminate this due to confusion over that instruction in making Version 8.

She also noticed that June’s difficult family relationships might’ve contributed to her relationship with Robert, and I wanted to make sure all players noticed that, so I made it clearer that June saw him as a father figure by saying it from the start in the synopsis. That way, the reader goes in knowing that, and they understand why throughout the story.

Finally, for the assessment (at 42:46), she said that she did learn about potential consequences and liked that even though Cerebra didn’t have malicious intentions, the technology did end up being used in malicious ways. She also felt the emotional part of the story through the discussion of the girls who were being tested on’s experiences, and she felt a positive connection with the protagonist. This was my first entirely positive assessment, so I felt good about the direction the game was headed in!

Version 6

I playtested with another female CS377G student (CS Coterm), with strong tech/tech ethics familiarity. She enjoys long-form fiction, making her part of my target audience.

She really liked the world-building (at 3:40), character-building (at 31:38), and inconsequential choices (31:40), all of which I’d been iterating on and/or debating extensively, so it was really great to hear. She suggested including more inconsequential choices (e.g. follow up questions that June can ask to Elena about the information Elena gives her). Butch had given guidance that inconsequential choices could be disheartening for players, so it was good to know that the way I implemented them resonated.

She liked how Robert’s character and his relationship with June were built up (at 39:28) and suggested using that for even more impact when Robert and June’s relationship fell apart. She also felt that the player didn’t get as much about Elena, which she recognized could be because Elena was a government official and her relationship with June was serious work (at 38:51), but it did encourage me to build up Elena and her relationship with June a bit more in the next version.

She also suggested switching the button to start the story to be “Start” instead of “Next” (the default for next page), which I couldn’t figure out how to do before Version 7, but I was able to for Version 8!

Finally, for the assessment (40:42), she had a similar answer to my first question (learning about the potential consequences of unchecked technology) as my Version 5 playtester: she thought it explored the topic well and also liked the concept that even though Cerebra’s employees thought it was being used for something, it was actually being used for other things. For the second question, she felt that reading through the ages of the girls being tested on gives emotional impact, but she suggested again building the relationship with Elena a bit more to make the fall out more impactful.

Version 7

I playtested Version 7 twice: with a CS 377G student who has a lot of tech, tech ethics, and creative writing experience (Player 1) as well as Christina (Player 2), who has lots of game design and writing experience. The video shown above is for my second playtest as there were some technical issues with getting the video from my first playtest uploaded.

Player 1 was my first playtester to actually interact with the game menu. My game used ChoiceScript’s built-in UI, and we found that some menu options didn’t apply, so I eliminated them. Player 1 also asked for a back button, so I implemented that (see Figures 6 and 7) but made sure it did not show up on pages with choices (so players couldn’t go back an alter their previous choices).

Figure 6: Back button on text-only page

Figure 7: No back button on choice page

Both Players 1 and 2 wanted more choices, and Player 1 specifically asked for a choice to break up the first flashback (a June/Robert scene) and for more choices directly interacting with Robert, so I implemented more choices, including a choice in all flashbacks with Robert.

Player 1 did ask that I restructure the story so that it played out across 2 years instead of a few days (with flashbacks), but I was hesitant to do so because I didn’t personally feel like it was necessary. To provide a point of comparison, I conducted my playtest with Player 2, and they did not feel that it was necessary so I kept my timeline as is.

Player 1 also found the characterization of June to be weak, so I added more internal reflection (for example, before/during/after her first weird interaction with Elena). Player 2 suggested making it a choice to go after Elena or not after that weird interaction, and even though both choices ultimately come back to the same storyline (i.e. are inconsequential), the player learns different things about June and gets to experience different story. Player 2 also didn’t have a problem with the existing inconsequential choices (at 15:24), providing more encouragement to continue implementing them.

Player 2 also really liked one of the later consequential choices (at 28:29), which was great feedback for me because it involves choosing between two characters, and earlier playtesters felt that I didn’t build up the relationships with those two characters (Robert (Version 4 feedback) and Elena (Version 6 feedback)) well-enough, so this indicates that there’s a strong pull on both ends, facilitated by strong characterization.

Player 2 also suggested paring down on the descriptive language but mentioned that she really enjoyed the writing (at 34:24). I made sure to cut out instances of it that weren’t as necessary, but as she acknowledged, it is a personal preference, and I do tend to like more descriptive stories. However, I did make sure to ask my next playtester how she felt about the description level.

Player 2 did feel that the speed-thru ending diluted it and that I should stick with the pacing of the rest of the story. This goes back to the debate I had after the Version 3 playtest as to whether I should give away the ending quickly or leave the players wanting more. This ultimately convinced me to end while still in the current pace/present tense (i.e. give the players a little but still leave them wanting more).

Finally, for the assessments, Player 1 already knew a lot about tech ethics, so they didn’t feel like they learned more, but they did feel emotional connection to the consequences through the story and said it could be strengthened by June pointing out that the girls being tested on are girls of color (something implied but not explicitly said). This assessment felt overall pretty good to me since it’s probably not possible to teach someone who already knows so much about the topic area—other players who didn’t know as much felt like they learned a lot—and I was happy that I was able to make that emotional connection, even for someone who does know a lot about the area.

Player 2 echoed Player 1’s assessment of the emotional aspect of the story, so I made sure to be explicit about the race dimension. They felt that some emotional aspects (e.g. calling the program “Operation Mindless” and mentioning explicitly that the human testing subjects were orphans were too on the nose, so I cut those out). Player 2 also liked that there is no “winning” ending (at 36:00) and suggested that I take it one step further by letting players play through other branches (i.e. go back to the most consequential decision they made and pick another path). I added this in to show players that we will always lose if we don’t make space for important ethical and moral discussions within technology.

Version 8

My final playtest was with the same friend who tested my Version 2 (CS Coterm, has taken tech ethics classes).

The playtest went smoothly for the most part, and she mentioned that she noticed that the riskiest option tended to go first, so I made sure to go back and vary the order of the choices. It was a similar piece of feedback to what I got in Version 5 about the choice order (felt pressured to select the first choice), but it gave a different action item (rearrange the order rather than make it clear that order doesn’t matter). I also took into account feedback from Version 7 (at 0:38 in that video) where the player was confused over instructions that order of choices doesn’t matter as well to ultimately drop the instruction but vary the order, which should partially address Version 5 feedback. After discussing with Butch, this decision also made sense because the order of choices not mattering goes as an assumption of the genre, and my Version 5 playtester might’ve just been unfamiliar.

A few earlier playtesters had also mentioned that the amount of text in a paragraph/on a page were too much, and (at 34:58) this playtester mentioned that it felt a lot easier to read than the first time she played, and she also didn’t mention having the issue mentioned before of too much description.

Furthermore, she also liked the jump-back ending that I added in response to Version 7 feedback (at 35:38), and one of her biggest pieces of feedback the first time playing was that the ending didn’t feel satisfying, so I asked her about it this time around, and she likes it now (at 36:12)!

Finally, for the assessment (at 32:31), she felt that the game met both objectives. Given the generally positive assessments of both my objectives and overall gameplay experience in my later playtests, I feel confident that Erosion is in a much better place when I started.

Reflection

I really enjoyed building Erosion. It tapped into issues I feel deeply about, and I got to combine my interests in game design and writing.

The biggest thing I learned was how to develop a game with many complex moving parts (i.e. branches). I came in a bit too confident and tried developing in ChoiceScript from the start, which was, as the teaching team had suggested, not the best move.

The first version I made was also not scoped out well—it was the start of a novel without a clear ending point, unlike my current story which is the start of a longer story but does have a clear ending—so I originally thought that just that was the issue instead of the unnecessary complexity also contributing to difficulties. For my next version, I brainstormed a lot and came up with a more doable storyline but stayed in ChoiceScript. It took me until my third version to finally break the story down into a simpler format (a flow chart). From there, I realized I had overwhelmed myself with choices and created a version of my game with just 2 choices, following a suggestion Butch made! After that version, it was much smoother sailing: I gradually refined and added in more choices until I got to the version I have currently.

Next time, I would start with scoping out the start, middle, and end of my game (instead of doing so after Version 1). Then, I’d determine 1-2 choices to add and implement them in a simple format to playtest. After that, I’d build up the choices I was using, either in the simple format or in my final one (i.e. ChoiceScript). This would’ve enabled me to spend more time refining rather than figuring out where to even start.

That said, I am very glad to have had the experience of pivoting so many times and still landing on a game I’m proud of because it taught me a lot about my design process and will set me up well for future projects!