Michelle Rundbaken and Yacine Merzouk are the creators behind Society of Curiosities, a tabletop puzzle game company. They offer a quarterly subscription box with a continuous narrative arc, starting with Madok’s Lost Treasure; standalone digital games, including The Bewitched Circus; and a series of lighter tabletop murder mysteries called The Fairytale Files.
As I’ve played through Society of Curiosities’ games over the past year, I’ve been consistently blown away by their artfully handcrafted physical components, engaging puzzle-driven storytelling, and immersive tech. Their deep approach to world building has transported me to intriguing alternate realities closely entwined with our own, and I’ve often wondered: what type of research goes into creating this style of gameplay? And just how deep do Society of Curiosities’ rabbit holes go? Curious to learn more, I opened a Zoom portal to Michelle and Yacine’s Hawaiian oasis for this interview.
Note: This article contains light spoilers for elements of Society of Curiosities’ games.
Conversation with Society of Curiosities
What were your backgrounds prior to escape rooms?
Yacine: I was a gamer my whole life, mostly RPGs. I did a bit of freelancing and self-publishing in the D&D space. And I’m a software engineer by trade.
Michelle: I have a background in psychology, education, theater, dance, and art.
You used to own Kauai Escape Room before switching to making tabletop puzzle games as Society of Curiosities. How does designing tabletop games compare to designing in-person escape rooms?
Michelle: Because we don’t have a time constraint in tabletop games, we can tell more story. We can add in more details that you just couldn’t do in an in-person room. We can do a little more on the writing side, too, though we’re still very conscious that people don’t like to read. And we can do anything we want because we’re not in person! We can make you travel to different countries. We can make you time travel. We can blow things up.
Yacine: In person, building the set is a ton of work. It would take anywhere from one to three months after the whole game was designed to just build it. But immersion is automatic. Although storytelling is easier in digital and tabletop games, it’s harder to set the scene and tone and get people to buy in.
Michelle: Also, you have a little more leeway with the set design in a physical escape room because people are moving so fast through different rooms and they don’t pay as much attention to all the little details. But when you send out a physical package, everything has to be spot on. You really have to up your design skills, your attention to every little detail, down to researching certain fonts that would be used in a specific period of time and hiring a person who can do that handwriting. In an in-person escape room, you can get away with just writing things yourself.
Who is your target audience for Society of Curiosities?
Yacine: The recommended age range is 14 and over. The most success we’ve had commercially with Society of Curiosities is the geek crowd, when we were mentioned in Ars Technica. That crowd is like “yes, I want to think in my free time.” Many people in that crowd are gamers. They have the imagination. They will buy into the story and fill in the blanks. They’re comfortable taking it and making it their own experience.
For Fairytale Files, we’re definitely going to a wider mass-market audience. We have a lot of families who’ve played with their kids, especially early teens.
Michelle: That was purposeful for us. Society of Curiosities definitely involves some level of skill, or some level of commitment, when you start. You don’t have to have the skills, but you definitely have to be willing to try and try again sometimes. Fairytale Files is intended to be much lighter — faster wins, fewer step puzzles, all of that.
How do you make the digital elements of your tabletop games not just feel like video games?
Yacine: For our digital interactions, you have to input your “answers” by talking to the characters in the game. It can never be something that you just click or a one-word answer. We want people to be able to talk naturally, and the AI has to be able to pick it up. So for example, in our Fairytale Files games, the map of Fairy is viewable online, but isn’t clickable on purpose. You have to tell your team on the ground “hey, can you go to the tavern?” It feels very different.
Michelle: We had someone early on tell us that they really appreciate that the items in our games are something you have to use. Not something that you see, not just a kitschy item that has a clue on it, because you don’t really need that, but items that you really use and manipulate for a game.
When playing your games, I’ve been impressed by the depth and realness of the historical materials. What does your research process look like?
Michelle: We do a lot of research early on. We choose a theme and then we start researching. And that informs a lot of our puzzles. The deeper we can get into a world, the more the puzzles come easily. If you are in a place, everything has a purpose, everything has a reason.
Yacine: For in-person escape rooms, people buy in quickly because of the immersive set. You can get the same effect in a tabletop experience if the initial intake gives you the right vibe. That’s probably where we want to be the most accurate or give the right feel of age and location. And once people buy in, you can have a little more leeway on historical accuracy if you want a puzzle that furthers the story. So we don’t have to be 100% all the time, but we do stay 95%.
What are some examples of small details that the casual player might never notice?
Yacine: People have more time to be nitpicky at home. In The Mysterious Map Heist, our digital game, you trace back something from the 1800s in France. Michelle went to an auction site, found a time-appropriate map, bought the digital rights…. and then someone emailed us after they played the game and was like, “That fort on the map was built 3 years after the game took place.” We left it in, but good on them.
Michelle: For the tattoo of a pirate, we hired a tattoo artist to make a ship tattoo, and it came back and had a marlin jumping and a mermaid at the front and it was stunning. But we had done a ton of research on pirate tattoos, and they were rudimentary. Even the ship we have now is too nice. This poor guy… he had to do three revisions, taking away from his beautiful art, until it was just a ship.
Yacine: The cursive in our games is also time appropriate. The cursive for the pirates game came back so beautifully from a calligrapher, and we were like, “This is way too pretty! This was written by a pirate on a ship. Can you just make it uglier?” And then the game from the ‘30s, The Memory Stone, features an archaeologist in a jungle, so Michelle found an expert on the Palmer method, the handwriting method taught in schools in the 1910s.
How do you approach character development in your games?
Michelle: We make really accessible characters who we design to be upbeat, happy, nonjudgmental cheerleaders. You have to work with them, and they’re who you’d want to work with.
Yacine: For example, in our first game, you’re supposed to be the brains, and there’s an ex-military guy, Steiner, on the ground. At first, people didn’t jive with Steiner — he was too serious — so we lightened him up a little bit, made him more of a lovable, bumbling idiot. In another game, Victor is a consultant who is hired when you unexpectedly end up in Europe. It was originally supposed to be a professor, but then players would defer to his authority because he was a professor. So we just changed him to a likable author character.
Exclamation points go a long way to set whether the character is upbeat or not! And you have to say most things in one sentence! It usually takes us 7 or 8 playtests before we hit the right tone or right objectives.
Michelle: You also need to have contrast — not every character in an experience can feel the same. As with escape rooms, you can use pretty iconic characters. They read well. You don’t need to have as many nuances because we try to keep down on the reading.
Do you have any tricks to help make your characters seem human over text?
Yacine: We have a timer of between 2 to 7 seconds to add a feel of realness, but it’s an illusion. Just a little delay so you don’t get a paragraph right away that indicates “robot”. When a game is new, our hardcore fans get the added advantage that we monitor the AI almost 24/7 for the first few weeks or months until it’s ready to be on its own. We get a text message every single time the AI doesn’t catch something automatically, and most of the time I will respond. By the time more casual players come in, the AI is good enough that they’ll get the same feeling.
Michelle: One challenge we have is when people don’t really buy in and they don’t communicate like the chatbot is actually a person. Those people will start sending just one-word responses, and the way the AI chatbot is programmed, it doesn’t really respond well to that. You have to have prompts when these things start happening like: “Hey, I’m not quite sure why you’re giving me one word. Can you give me a little more?”
Yacine: When people start being monosyllabic, it’s probably when the AI fails the most often. The strength of our games is that you can talk normally.
That’s so impressive. Can you tell me a bit more about the tech that powers these natural language interactions?
Yacine: I build everything from scratch! We have two tech stacks. One is for text messages only, which works really well in the US but wasn’t so hot internationally, so we also added a second stack for web-based messaging.
The SMS stack is just straight-up PHP. It builds on lexical fields databases that are available for free from universities. They give you a bunch of words with their synonyms or close variants. I also built my own lexical fields to supplement the databases. For example, in The Memory Stone, you have to push stones with ancient glyphs — that we actually had designed by a Native American artist so they were in the correct style. But because the drawings are not all obvious, we made sure that something like “four-legged thing” is a correct substitute for “dog” or “jaguar” or “cat.”
From there, I do a lot of work with game state and context. Part of speaking like a human is knowing where you are in the game. Context matters, especially when people use words like “it” or “that”. So “take it” in context could mean “take the shovel.” Another part of the context is if somebody is probably right but they haven’t quite proven it, the AI just needs to say, “I think you’re on the right track. Can you tell me more about what you’re thinking?”
What’s a problem you’ve encountered that someone might not anticipate?
Yacine: I do a lot of work on character strings (sequences of letters and numbers.) For example, we might be expecting a number as the solution to a puzzle, but we should also accept “the code is 1234” or “one then two then three then four” or “1-2-3-4.” I build my own little libraries that break down strings a dozen different ways and look for things in order. I also work with parts of a word so we can recognize the answer even if there’s a typo in what the player wrote. Let’s say you have a word like “excavation.” The group of letters “exc” or “cav” isn’t going to be in any other word for the context we’re in. If the AI doesn’t know how to respond to an obvious typo, it really takes you out of the game.
On that note, what AI are you using?
Yacine: I also built it myself! We don’t use a premade chatbot system like a messenger bot.
We started with SMS only. For programmers out there, Twilio does a great job of providing a way for people to send text messages. Then getting the Twilio system to talk to the web and keep track of the answers was challenging. We use a stack called Meteor, which does automatic updates of the webpage. Around 10% of people switch to the web halfway through the game, so we make sure game state syncs between SMS and web. Probably 80% of people choose web-based messaging over text messaging right off the bat.
You also make a TON of different websites for your games! How do you keep track of them all?
Yacine: This is my real expertise. I worked as a web developer for a marketing company for 7 years, and then I had my own web design and software development shop. So building a website for me is where I have the most tricks. I already had the expertise to build websites from scratch. I also did a lot of work with WordPress, so if it’s faster to do it with WordPress, I’ll do that. I have my own server with a ton of storage, so I can build as many websites as I want and we don’t have to pay more for each. If it fits the story, we can just do it.
The other thing is SEO. If I build a website, I’m fairly confident it’ll be picked up within a week or two and rank first. But when we absolutely need to rank first for some things, we need to find a word that nobody else uses. So for example, we sometimes come up with the name for a character which players will have to search for. If the search results have fewer than 8-10k results, essentially nobody is searching for it and you can confidently build a website that Google will pick it up in a week or two. Though we had one time where the website didn’t pick up by when the game was due to launch. So we bought ads for this thing that nobody searches for. You pay per click, so cents every month, and it boosts the ranking.
You’re constantly iterating and adding new fancy features to your games. What’s next?
Michelle: With Fairytale Files, we started giving hints within the messaging system. So for the next season of Society of Curiosities, we’ll do that, too. You won’t have to leave the game for hints.
Yacine: Fairytale Files is an easier difficulty level, so there are fewer hints that lead to the solution. There are like 3 hints and then it jumps to the solution. For the first Fairytale Files, we don’t show you the whole list of hints for the whole game, just what’s pertinent to where you are in the game. The next step is going to be you talking to the character to say you need help, and then the character, in their voice, nudges you in the right direction. That’s more challenging to build than it seems.
Michelle: The bottom line is that hints are there to get you unstuck. They should never be hard to access.
One final question: do you have any parting advice for players of your games or folks looking to create games in this style?
Michelle and Yacine: For our players, first of all, thank you! It means so much to us to have players to create for. As far as advice for the games: if you trust the game and trust us, you’ll have a much smoother experience. We promise your team on the ground is not withholding information from you. Also, take the time to take in everything. We love tabletop games because there are no time constraints — so take your time, and have fun!
For creators: immerse yourself in the world of tabletop and online games. See what you like and what areas you feel have room for improvement. Keep a journal of puzzle ideas — we often get the best ideas from puzzles that we don’t like that inspire me to make a “fixed” version. And lastly, after you create, test, test, test. Any area that you see players having problems should probably be tweaked.