I wake up and hug my dog, feeling his fur against my hands and the way he looks at me like I’m the best thing that happened to him. I carry him outside and let him run, then head into the chicken coop. There are eggs on the ground. I pick them up one by one, put them in my bag, and give each chicken a hug. They lean in. I grab some feed and watch them eat. That’s enough to make me feel good about the day.

These are some of my favorite moments in Story of Seasons. I do whatever I feel like, and the effort always pays off. The animals never let me down. There’s something about that steady rhythm that feels calm and safe in a way that’s hard to explain.

The biggest reason I love these games is that they give me back a sense of control. That’s something real life takes away pretty easily. In relationships, you end up compromising more than you’d like. At work, you take on jobs you don’t want, deal with people who drain you, and act fine about it. Life moves fast and doesn’t wait. These games give me space to breathe and hand the control back to me.

So how do they do it?

The foundation is a stable daily routine and has no real fail state. Every day runs on the same 24 hours. What changes are the weather, the season, maybe your plans? Crops didn’t work out this season? Try again next season. Didn’t find the ore you wanted today? Come back tomorrow. On top of that stability, players have the freedom and confidence to make real choices about the only resource that actually needs managing: time.

So you have full control and all the time you need. Now what? If the game just handed you a list of things to do and told you to pick some, it would get boring fast when everything is laid out and predictable, even a world where effort pays off starts to feel like a chore.

But this game goes the other way. It lets players discover what’s possible on their own.

I had no idea the mines had multiple floors. Nothing in the game told me. Then one day, while digging, I uncovered a staircase by accident. I was genuinely surprised. I spent a moment worrying about what was waiting below, whether I could handle it. Eventually, I went down and noticed the ores were starting to change. Different types, rarer materials. No NPC told me this. No quest sent me there. But I started coming back every day, thinking about how to use my time better, upgrading my pickaxe, and collecting supplies that would let me go deeper. All of it was my own call.

The game gave me control again.

Finding something on your own, through your own curiosity, makes the world feel bigger than it first seemed. And because the choice to keep going is always yours, the sense of agency stays with you.

There’s one more thing worth talking about: this world responds to what you do.

Hugging your dog, feeding your chickens, brushing your horse, these aren’t just cute moments. The animals respond consistently, and their affection builds over time. One of my favorite parts is being able to bring my favorite chicken to a competition. It made everything feel real, like the effort I put in actually meant something and didn’t just disappear as a stone dropped in water.

The same goes for the people in town. NPCs aren’t static. They don’t just pull random lines from a script. Your relationship with them changes, and as it does, so does how they talk to you. This is part of what drives players to take on one of the game’s quieter challenges: marrying an NPC. But it goes beyond that. Every character has their own story, shaped by time, location, weather, and season. The more you talk to them, the more you understand them, and slowly you build something that feels real. The world notices you’re there, and that makes you feel like your presence actually matters.

This game makes you feel important. Your choices matter. You decide what to do with your time. And every piece of the design points toward the same idea: come rest here, wander around, pet the animals, find your own goals. The sense of control is yours.

Strangely, I never thought about how this genre gets made, until I listened to a talk at GDC.

A Problem That Sounds Abstract but Is Actually Fatal

The speaker is the narrative designer of Ghost of Yotei. She shared her experience solving a problem they encountered during early development: in playtests, players were broadly confused about who they are, what their relationship with the world is, and why they are here.

This sounds abstract, but it is a fatal problem. Because no matter how big or how free the open world is, once players don’t know who they are, what they should do, or why they are here, the entire world feels less explorable and more hollow.

The First Hour Does Everything

The way they solved this was through a clear design goal: in the first hour of play, establish the player’s identity, their major purpose, and the world’s story. And the way to do this is not writing an even longer narrative, but rethinking the quality and quantity of content, and where it is placed.

Three Dimensions for Evaluating Content

The speaker used three key dimensions for evaluating content:

1. Quantity: Is the content too much or not enough?

I fell into this trap in my own MR multiplayer project. Initially, we let a single NPC tell a long story about the purpose, the world, and who the players are. The result: players zoned out. Not because they didn’t want to listen, but because it was just too much. Even worse, the key information got completely ignored. It was, at the same time, too much volume and not enough clarity on what actually mattered.

Later on, we split the information into different sections and gradually delivered it to the player. At the beginning, the NPC only tells players who they are and what to do now. Later NPCs explain more about the world building and the details of the quest.

Jesse’s feedback also helped us a lot. He mentioned that players usually only remember the last few sentences, and they respond to verbs. That made us adjust the NPC dialogue further, for example: “Go check that table, you should find what you need.” Letting the narrative lead to action, not just conveying information.

2. Shift and Reframe Verbs

You don’t need to invent new actions. You can let the same action mean something different in a new context. A slash can fight an enemy, clear an obstacle, or become part of cooking. The action doesn’t change. The experience does. Players don’t have to relearn anything, but they still feel something fresh.

3. Value: every piece of content should have a reason to exist.

No matter if it explains the story, provides world building, teaches a mechanic, or leads players to the next area, every piece of content a player encounters should carry useful information. Only when the purpose of that content is established can a designer evaluate whether it is actually serving that purpose.

Where You Put It Is as Important as What You Put

Having the content is one thing. Knowing how to arrange it is another. Their team used heatmaps to analyze player paths, find which areas were popular and which were not, and used that to decide what content to place where.

Use Player’s Starting Point

Even in an open world, the player’s entry point can still be designed. This made me think of Breath of the Wild: you wake up in a cave, learn the basics, and when you step outside you’re standing at a high point. You can see the whole world from there, and your final goal, Ganon, is visible in the distance. One scene gives you the controls, the motivation to explore, and your major purpose. Nothing is forced, but you already know what to do.

Distance as a Teaching Tool

Another key idea is physical distance. If some content is complicated and requires a lot of prior knowledge, it shouldn’t appear too early. You can place it far from the player’s entry point, and plant NPCs, environmental cues, and quests along the way. By the time the player actually encounters that content, they are ready.

Spreadsheets and Content Maps

At the practical execution stage, their team used spreadsheets alongside content placement maps to ensure every piece of content provides key context and has a logged position on the map. I find this approach very useful. It allows designers to check whether players can meet the right content at the right location at the right time.

When, What, Where

The talk concluded with three key questions about content: when to give it, what to give, and where to give it. Align these three things, and players won’t easily get lost.

One extended suggestion I also agree with: after players leave the opening area, you should keep doing onboarding, but not to teach them how to play. Instead, teach them what is special about this new area, what they can do here, and why it matters in the world.

Carefully Crafted Freedom

This talk made me realize that the freedom I feel in an open world is not just about putting everything in the world randomly. It is designers carefully arranging and planning everything you see, making you feel like you discovered it yourself, like every decision was yours. That is actually a much harder design challenge than a linear experience.

Walking into GDC this year felt like standing at the center of a gold rush. Veterans and newcomers alike were packed shoulder-to-shoulder outside every session hall, craning their necks to catch a glimpse of what’s coming next. AI talks were everywhere — topics that barely came up last year now dominated nearly every corner of the conference.

So what did I actually take away? And more importantly, what does this mean for the future of game studios — and for players?

Everyone’s Racing to Strike Gold

The scale of investment big tech is pouring into AI is staggering. From automating documentation and streamlining development pipelines to generating assets and producing entire terrains in real time, the tools are getting faster and smarter at a pace that makes it hard to justify not using them. Multiple speakers mentioned seeing development speed increase by three to four times after adopting AI. For studios already stretched thin on resources, this isn’t just a convenience. It’s a structural shift.

Honestly, I used to be skeptical of AI — resistant, even. I’ve come to realize that resistance was mostly fear: fear that AI would take jobs, including mine. But after sitting through session after session at GDC, my perspective started to shift. Not a single speaker suggested that AI would make designers obsolete. If anything, every conversation was about how to work with AI.

That realization hit me differently. People with strong fundamentals are exactly what companies need most right now. AI is a powerful tool — like a sharp knife. Whether you’re an experienced chef or a first-time cook, that difference still shows. Jesse shared a similar observation after his talk: a seasoned Art Director tends to get more out of AI than a junior artist, because they understand the underlying principles well enough to direct it effectively. They iterate faster, and they iterate smarter.

The Shovel Can’t Mine on Its Own

What stuck with me just as much, though, were the sessions about what AI still can’t do.

Two separate talks arrived at the same conclusion: AI cannot, without human intervention, independently develop a game. Jesse shared one particularly striking example — the same codebase produced wildly different results across two playtests. The reason? Without clear rules and structure, AI behavior becomes unpredictable. In Google Cloud’s session “How AI is Reshaping Game Creation,” a speaker from Unity made the same point: when AI develops without any human oversight, the output tends to fall apart. It starts breaking its own rules, and the original design intent evaporates entirely.

This matters a lot. A game is, at its core, a rule system that creates meaning. An AI that violates its own rules isn’t just creating bugs — it’s dismantling the entire experience.

So, Where Does That Leave Us?

Google’s framework in that same session categorized AI’s value by studio size: small indie studios can use AI to fill resource gaps and venture outside their core expertise; mid-sized AA studios can treat it as an innovation accelerator; large AAA studios can embed it as an intelligence layer to maintain quality at scale.

It’s a reasonable framework, but it feels a bit rigid to me. The most compelling uses of AI aren’t exclusive to big studios. It was never really about how large your team is — it’s about what your players actually want.

The question Jesse raised after his talk has stayed with me: What games could only exist because of AI? That’s the space worth exploring — not just using AI to do the same things faster, but discovering what creative territory it opens up that simply wasn’t possible before.

And to get there, the key skill is learning how to direct AI. Design intuition and hands-on experience are the real currency of this era. Everyone now has access to a tireless, highly capable assistant — but you still have to know how to delegate, how to give direction, and how to push it toward something meaningful.

When AI Becomes the New Normal

Smart NPCs are the most obvious near-future AI application — characters that genuinely respond to players, remember decisions, and shift behavior accordingly. It makes me think of a metaphor: if the game of the future is a stage built by designers, then AI is the cast of actors. Designers set the rules of the world and the outline of the story; AI improvises within that framework, performing a unique show for every player. The designer’s role doesn’t disappear — it evolves from puppeteer to playwright and director.

That’s exciting. But novelty fades, studios will catch up, and players will adapt. A year or two from now, it’ll just be another expected feature. So then what?

That brings me back to a more fundamental question: What do people actually need from games?

Maslow’s hierarchy of needs turns out to be surprisingly relevant here. As AI generates increasingly personalized and seamless content, will players crave something AI can’t simulate — a real, human connection? The rise of social media has already eroded our opportunities to practice genuine interaction. The resurgence of tabletop gaming, or the enduring popularity of community-driven games like Minecraft, might all be pointing to the same thing: what players sometimes want isn’t better technology. It’s a real reason to be with other people.

The games that truly matter may not be the ones with the most advanced AI — but the ones that best understand what players are longing for right now.

Faster Iteration Means Listening Harder

One theme kept resurfacing among successful studios: build your community, then actually listen to them. AI dramatically accelerates development iteration, which means the ability to extract genuine signal from player feedback becomes even more critical, because now you need to make the right calls faster.

It’s not just about collecting data. It’s about knowing how to filter it: which voices carry real insight? Which emotional reactions deserve serious attention? AI’s strength in processing large volumes of feedback can help surface patterns quickly, but deciding what actually matters for the player experience still requires human judgment. Figure out your product goals first, then decide what’s worth tracking — not the other way around.

Give Players What They’re Looking For

Looking back on GDC, every conversation about AI kept circling back to the same question: What do players actually want? No matter how powerful the tools get, a game that doesn’t resonate is just wasted effort.

The best games of the AI era will probably look something like this: a robust set of rules and systems that give the game its foundation, with enough room for AI to generate something unique for every player. But where does that foundation come from? There’s only one answer — you have to genuinely understand your players first, then design the game they deserve.

The gold rush is real. But the designers who find their footing in this wave won’t just be the ones digging fastest. They’ll be the ones who stop to ask: What exactly are we looking for?

Ace Attorney is structured around two distinct phases: the investigation phase and the courtroom phase. During the investigation phase, players explore different locations and solve puzzles — but each location is presented as a single, fixed-perspective screen. This fixed framing ensures that players search only within what the designer has already decided they should see.

In VR, that frame disappears entirely. Players can look anywhere, at any time, and you simply cannot force their attention.

When the Screen Disappears

A fully immersive, room-scale puzzle experiences fascinate me in a way that’s hard to describe, which is part of why I wated to dig into how the existing ones actually work. This research looks at three games, A Fisherman’s Tale, Ghost Giant, and I Expect You to Die.

Case Study 1: A Fisherman’s Tale — Layering Sensory Cues Across Scale

A Fisherman’s Tale is built around a Russian doll mechanic: the player stands inside a fisherman’s hut, which contains a miniature model of the same hut, which contains an even smaller model inside that, and so on — each layer mirroring and influencing the others. The scale mechanic itself is already a form of indirect control. Rather than using UI elements to signal what’s important, the game uses the inherent strangeness of proportion to make the player curious.

A Voice That Sounds Like Your Own Conscience

This game has an interesting voice-over. There’s a voice that speaks to you constantly, but it never gives you answers. Instead, it speaks the way your own internal monologue might — guiding your reasoning rather than your actions. When you’re stuck trying to find a key, it says something like: “It’s locked — it must have been put somewhere last time it was used.” That line doesn’t tell you where the key is. It nudges you to think about where you would have left it if you were the fisherman. The player and the character become one, which keeps the immersion intact while still functioning as a hint system.

Planting a Memory Before You Need It

One puzzle stood out to me as an example of long-form indirect control, the fishhook puzzle. Early in the game, the player encounters a fishhook and interacts with it — nothing major, just enough to register it in memory. Later, when a locked door becomes an obstacle, the solution turns out to be the fishhook itself. But because of the scale mechanic, you need a key that matches the size of the lock, which means you have to retrieve the hook from inside the miniature hut.

While I can appreciate the intended effect, designing this kind of associative puzzle requires a clear awareness of one fundamental constraint: players have limited cognitive load. The more information competing for their attention, the less likely they are to retain the connections that matter. A few things are worth keeping in mind:

1. Temporal and Spatial Distance
You don’t want players to encounter puzzle-relevant information while being simultaneously bombarded by other stimuli — if their cognitive load is already at capacity, the hint simply won’t stick. The same principle applies to physical space: if the answer and the puzzle are too far apart, players may walk right past the connection and never make it. Cognitive load isn’t just about how much information exists — it’s about how much the player has to hold in their head at once, across both time and space.

2. The Number of Logical Layers Between the Puzzle and Its Solution
Each additional layer of logic stacked between a clue and its answer adds to the player’s cognitive load. The fishhook puzzle is a good example: recognizing that a fishhook can function as a key already requires an unconventional leap in thinking. Introducing a second layer — that the fishhook must be of a specific size, retrieved from a miniature model — compounds that demand significantly. The voice-over can help bridge the gap, but beyond a certain threshold, no amount of hinting can compensate for a chain of logic that has simply become too long for most players to follow without losing the thread.

Audio and Color as Guidance Tools

VR has a distinct advantage when it comes to spatial audio. Unlike flat-screen games, VR can anchor sound to specific points in three-dimensional space, making audio cues feel less like a game mechanic and more like a natural property of the world — which in turn drives attention and curiosity without breaking immersion. Color guidance, while common across all game formats including PC, remains equally valid in VR and should not be overlooked simply because it isn’t platform-exclusive.

A Fisherman’s Tale demonstrates how these two cues can work together. In the lobster shell puzzle, the game never directly tells you that a key is hidden inside. Instead, it offers two independent signals: sound — shaking the shell produces a faint rattle — and color — a subtle green glow visible on the interior of the shell. Either cue alone is sufficient to guide an attentive player. Layering both ensures that almost no one gets permanently stuck, while keeping the feeling of discovery intact.

Water Pipes: Teaching a Mental Model Without Explaining It

The water pipe puzzle is worth noting not for its complexity, but for what it reveals about the game’s tutorial design. The puzzle uses only straight pipes — no bends — and this feels intentional. By this point in the game, players have already been trained through earlier puzzles to look for color and size as the primary solving dimensions. When the pipe puzzle arrives, experienced players will scan for color matches and compatible sizes almost automatically, because the game has quietly built that habit into them. And once internalized, it does something more valuable than any UI marker: it drives the player’s attention to exactly the right place, without them ever noticing they were guided there.

Snap Mechanics: Removing Noise, Focusing Attention

In VR, players can manipulate objects from any angle and orientation, which is both a feature and a design challenge. If a puzzle simultaneously demands that the player find the right object and orient it correctly, there’s a real risk of the player getting stuck on the wrong layer of the problem.

A Fisherman’s Tale handles this by separating concerns. Puzzles generally ask for one thing at a time: either find the correct object and snap it into a slot, or solve a directional/rotational challenge. When orientation matters, the object is usually snapped into position first, and then the directional element is introduced. This keeps the player’s cognitive load focused exactly where the designer intends, rather than scattered across multiple simultaneous problems.

The Setup Before the Light

Many VR games make use of spotlight lighting to draw the player’s attention, and A Fisherman’s Tale is no exception — but what I find more impressive isn’t the glowing radio itself. It’s the sequence of visual guidance the designer constructs before that moment ever arrives.
First, the designer leads you toward the south side of the room to look outside, introducing you to the Russian doll concept. Then you’re turned 180 degrees to face the miniature lighthouse, where you’re invited to interact with it — and to reckon with your own giant hand reaching in from outside. Only after all of that does the moment arrive: you look up, and the radio is right there, the only thing in the room that’s lit.

Case Study 2: Ghost Giant — Emotional Attachment as a Gaze Anchor

Ghost Giant takes a different approach to indirect control — one rooted not in environmental design or physical mechanics, but in emotional investment. The game opens by establishing something simple and powerful: you are a giant, and there is a small creature that needs your help. That premise does an enormous amount of work.

The NPC as a Continuously Moving Focal Point

Once you care about the main character — and the game makes sure you do from the very first moments — guiding your attention becomes almost effortless. Your eyes follow the character naturally. When he looks at something, you look too. When he moves, your gaze tracks him. The character becomes a living visual anchor that the game can move around the world to direct your attention wherever it needs to go.

This is, in my opinion, one of the most elegant indirect control strategies available in VR: if you can make the player genuinely care about a character, you get gaze control almost for free.

Establishing the Player’s Identity Through Early Interaction

The opening sequences — helping build and stabilize a small structure — aren’t just tutorial moments. They’re also communicating something more fundamental: what you are, what your scale is relative to this world, and what you’re capable of. The player internalizes their own identity and agency through these interactions, which makes subsequent guidance feel grounded rather than arbitrary. Light and sound cues layer on top of this, signaling progression and pointing toward next steps, but they work because the identity foundation is already in place.

Case Study 3: I Expect You to Die — Cinematic Openings and the Joy of Feeling Smart

I Expect You to Die does something I haven’t quite seen done as effectively in any other VR puzzle game: its opening sequence is genuinely cinematic in a way that uses the medium’s unique properties rather than fighting against them.

The Opening: A Camera Rig You’re Sitting In

The intro turns your chair into a camera dolly. You can still look wherever you want — this is VR, after all — but the visual design is so relentlessly engaging that you almost don’t want to look away. Rapid shifts in light and shadow, dramatic changes in scale, strong leading lines, particle effects and visual punctuation — all of it is competing for your attention in the best possible way, refreshing your visual interest faster than you can settle into any single point.

The tool I found most effective here is scale contrast. When something enormous appears beside you, your eyes pull toward the distant mass. When you’re enclosed in a tight, narrow space, your instinct is to look for the exit. The opening uses both of these primitive attention responses in sequence, and the result is a kind of visual rollercoaster that guides your gaze through a narrative without ever asking you to look at anything specific.

Inside the Puzzles: Interesting Objects Over Glowing Indicators

Once you’re into the actual gameplay, I Expect You to Die largely abandons obvious light-based guidance. Instead, it draws your attention through inherently interesting objects — a cocktail recipe written on a scrap of paper, a robot companion that reacts to your actions, a colleague speaking to you through an earpiece as you work through a crisis together. These elements earn your attention because they’re engaging on their own terms, not because they’re emitting a visual signal.

The puzzles themselves are built on spatial logic and creative constraint. Objects you need are often just out of reach, requiring lateral thinking about what tools you actually have available. A leaking spacecraft capsule needs to be sealed — you can use the cigar that’s been sitting in your environment the whole time, or the space food pouch you find in an overhead compartment. Multiple valid solutions exist, which accomplishes two things simultaneously: it makes the player feel clever when they find any solution, and it reduces the chance of total frustration when one approach doesn’t work.

That last point deserves emphasis. Making the player feel smart is not incidental — it is central to the design. The moment a player feels like they outsmarted the game, rather than being led through it, is the moment they’re motivated to push into the next challenge.

Conclusion

Indirect control in VR isn’t just about applying existing design principles — it demands a finer level of craft to guide the player’s gaze and sustain their curiosity in a medium where nothing can be taken for granted. Here’s what I took away from these three games:

The first is to lean into every sensory channel available. Spatial audio, voice-over, color contrast, visual effects — VR has tools that flat-screen games simply don’t, and the best puzzle games use them not just as guidance mechanisms, but as a way to make the world feel genuinely alive. That said, layering these cues requires careful consideration of cognitive load: too many signals at once, or hints planted too far from the puzzle they belong to in time or space, and the player stops absorbing and starts drowning.

The second is to make the player care. When you’re invested in something — a character, a story, a relationship — your attention follows naturally and without resistance. Building an emotional connection between the player and the world isn’t just good storytelling; it’s one of the most powerful and invisible guidance tools a designer has. Getting the player to care means you don’t have to fight for their attention anymore. It comes to you.

VR puzzle design is harder than its flat-screen equivalent precisely because the designer can never assume control of the player’s eyes. But that constraint, when taken seriously, pushes designers toward solutions that feel more natural, more immersive, and ultimately more satisfying — because the player always feels like they got there on their own.

References

[1] “Phoenix Wright Ace Attorney: Episode 1 — The First Turnabout”, Nuch, YouTube, 2024/04/05. https://youtube.com/watch?v=9DhB0ALuAVM

[2] “A Fisherman’s Tale Oculus Quest Gameplay”, BMFVR, YouTube, 2020/07/20. https://www.youtube.com/watch?v=ZBJiGeUX73Q

[3] “Oculus Quest Ghost Giant A Heartfelt Emotional VR Game”, BMFVR, YouTube, 2020/02/19. https://www.youtube.com/watch?v=WPTQSGNAagg

My inspiration came from the board game Saboteur, which is also a path-building game, but… wait, is it really?

Let me briefly introduce Saboteur. Players are divided into two factions: good dwarves and bad dwarves. The good dwarves’ goal is to build a path to the target mining spot and win by finding gold. The bad dwarves aim to prevent the good dwarves from achieving their goal. Each turn, players have three choices: lay a path, use an action card, or discard a card, then draw one card from the deck. Because the deck has a limited number of cards, if the good guys haven’t found the gold after all players have used their hand cards, the bad guys win. There are always fewer bad guys than good guys in the game.

What Makes It Tick?

1. Hidden Factions

I think this is the biggest highlight. Because you have hidden allegiances, every player’s every move represents crucial information. You must “think” and “pay attention” every turn, creating uninterrupted turn-based gameplay. Even when it’s not your turn, you must try to decode the true intentions behind each player’s actions.

2. Scheming

This naturally emerges from the first point and the game’s inherent constraints. Since villains are always outnumbered by heroes, villain players must hide their intentions or have no chance of winning. You must “take a risk” — pretend to be a good guy helping build the path. This action threatens your own victory, but you have no choice. Only when the deck is almost depleted can you safely reveal yourself.

3. Theme

The theme allows players to immerse themselves. Action cards, path cards, and role cards all align with the theme — you won’t suddenly see a UFO blow up the entire mine and break the atmosphere.

4. Party Game

Everyone enjoys it together. Even spectators can guess who the bad dwarves are. This communal nature adds social value to the game.

Being the Bad Guy Is Hard, But Fun

Poker face management is really important. The moment before you flip your role card, the game has already begun. If you get a bad dwarf, every subsequent move requires pretending to be a good guy to avoid getting caught too early and failing completely. My own method is to temporarily adopt the good guy’s thinking pattern — even my words have to go through the “good guy brain” first. This has resulted in several occasions where I genuinely forgot I was the bad guy and got scolded by my teammates for being a game breaker.

Saboteur is really about playing hidden identities and faction warfare. Many other similar games, like Avalon and Werewolf, share this core concept. When you must attend to both the surface-level task and your hidden private mission, it adds tremendous psychological depth to the game, challenging your skill at disguising true intentions. This skill differs from athletic ability or intelligence — it involves managing every medium through which you might leak information: eye contact, facial expressions, body language, tone of voice — everything must be calibrated perfectly. And when you endure until the last moment, certain of victory, that dramatic evil laugh, or when you’re backed into a corner and forced to reveal your cards for a final showdown — these moments of information revelation bring joy and climax to the game.

Letting Players Be Good Actors

Of course, all these fun elements require perfect setup. Being forced to reveal too early, or not needing to hide at all, directly destroys the game’s enjoyment. It’s like drama or movies — if you know who the villain is from the start, wouldn’t that be boring? The design of this game’s path and action cards all help players “perform well” in this play.

Card Shape Reveals Intent

Cards are rectangular, meaning the path cards you get limit the directions you can turn. I only discovered this when designing my own path cards — square path cards make direction meaningless. When you notice someone laying paths in an unreasonable direction, it’s time to start a conversation.

Hidden Hand Cards, Trust Capital

Of course, you can argue that you simply don’t have the right path card in your hand. Your claim has credibility because your hand is hidden.

Hidden Gold Mine, Revealed Credibility

There are three goal cards in the game, and only one is the gold mine, meaning worst case you guess wrong on the first two. The “map” action card occupies a special position in the game — its effect lets the user peek at any goal card. This gives both good and bad factions a chance to perform. You can claim you’re a good guy and that the second goal card is the gold mine — everyone dig that way! Or you can say the second goal card isn’t the gold mine, so everyone should proceed to the other two locations. How to interpret revealed information depends entirely on the small clues you’ve collected.

Unequal Tools, Yet Balanced Gameplay

When you suspect someone is a bad guy, you need means to stop them. And if there’s stopping, there must be rescuing. The balance between the three types of sabotage tools is also interesting. A player can have up to three different types of restrictions placed on them, and rescue cards can remove one type or even two types at once. This design gives you the possibility that after two players restrict you, one ally can rescue you and you can continue sabotaging paths — providing balance for the outnumbered villain faction. However, this also depends on players’ relative positions, so sometimes it comes down to whether you and your teammates are lucky enough in positioning.

The Existence of Ultimate Moves, Reversals Upon Reversals

Of course, when you think the villains are at their end, if they pull off a comeback from the brink, that’s a reversal of reversals — absolutely brilliant. The ultimate card that destroys existing paths adds dramatic effect to this play, but the prerequisite is getting this rare card, being a bad guy, and having no restriction cards in front of you.

What Can Be Borrowed?

So Saboteur’s essence isn’t about digging tunnels. Is the game I want to create really about building paths? Saboteur’s various designs all revolve around the core: hidden identities and faction warfare. Under the simple, clear objective of finding gold, it adds various methods of hiding information and means of revealing it, and these elements and mechanics all reinforce its core.

My game actually leans more toward strategies for dealing with uncertainty, without wanting to hide factions. Perhaps I should focus on how to introduce uncertainty, and players should also have corresponding means to face and handle uncertainty. This would strengthen players’ sense of achievement after overcoming challenges, rather than relying purely on luck like Snakes & Ladders, where players have no way to intervene.

You might call this a puzzle game, but earlier puzzle-solving games felt more aligned with what I imagined that genre to be. Their core mechanic was about making a single choice to help an in-game character overcome an immediate crisis. For example, a character might be about to drown, but if you place a pole in the right position, you redirect the water flow and save them. Various iterations of similar gameplay followed.

The difference? The older games challenged your problem-solving skills. The newer ones challenge your impulse control.

This observation, combined with discussions about game design and the rapid development of AI, has led me to a troubling question: is the future of gaming heading toward addiction-driven design?

The Industry’s Open Secret

A while ago, wanting to better understand game design, I started watching YouTube channels dedicated to the subject. In one episode — which I unfortunately can’t find now — the discussion centered on how game companies, especially mobile game developers, rapidly release games to sustain their revenue. The video explored how these games are designed to quickly capture player attention rather than featuring deeper design elements like story, characters, or complex mechanics.

Trust me, I tried to find the video, but I really failed. However, I found a GDC talk[1] that is relevant to the topic I want to explore. In that presentation, you can see that producers focus heavily on various revenue calculations and analyses. The primary purpose of collecting player data isn’t to make games more fun — it’s to figure out how to make more money.

Take King, the developer behind Candy Crush, as an example. They operate like a game factory, releasing dozens of similar-mechanic games each year to test which ones stick with players. The successful ones stay; the failures are quietly shut down. This “throw everything at the wall” approach prioritizes scalable monetization over creative innovation.

Companies need to survive, after all, and this has become an industry standard. They inevitably hire talented people to design these kinds of games, even if the designers themselves may not enjoy doing so.

When AI Learns to Addict

With the rise of AI in recent years and its proficiency in analyzing human behavior — as evidenced by countless news stories about people asking AI for psychological advice or relationship guidance — it’s clear that AI’s growth potential could exceed our imagination. But what happens when this technology is deployed not to help us, but to hook us?

Applications in the gaming industry have been around for years, with numerous patents and technical papers published. These include EA’s Dynamic Difficulty Adjustment (DDA) patent, which uses AI and machine learning through the cloud to adjust game difficulty with the goal of increasing player engagement[2][3][4]; Zynga’s AutoModel system, which uses deep reinforcement learning to personalize player experiences[5][6]; and various academic papers exploring how to analyze and predict player churn[7][8][9].

These developments raise a concerning question: what happens when AI is fully deployed to analyze data and hook users with the goal of maximizing addiction? This brings me to another deeply concerning issue: the age at which people first encounter digital devices is dropping rapidly. When young children — whose brains are still developing impulse control and decision-making capabilities — encounter these highly addictive games designed by teams of psychologists and optimized by AI, what happens? We’re running an uncontrolled experiment on an entire generation.

The Creative Dilemma — And What Designers Can Do About It

As a game developer, I believe the vast majority of designers want their work to be meaningful. But if a successful free-to-play (F2P) gacha game has a replicable monetization model, can the balance between creative vision and commercial pressure really be maintained?

After attending GDC this year, I think the answer is yes — but it requires an intentional choice.

The most important thing I took away from GDC is this: knowing how to use a tool is not the same as knowing how to use it well. AI can accelerate development by three to four times, generate assets, and optimize pipelines — but a designer with strong foundational knowledge will always use these tools more effectively than someone chasing shortcuts. An experienced Art Director using AI will outperform a novice not because of the tool itself, but because of the judgment behind it.

This means the real question for designers isn’t “should I use AI?” but “what am I using AI for?” There’s a meaningful difference between using AI to engineer more addictive reward loops and using it to build richer, more personalized player experiences. Caryl Shaw’s Two Dots is a useful example: it’s still F2P, but its payment mechanics are built around the motivation of helping friends — players pay not for their own progress, but to assist friends who are stuck. This transforms competition into cooperation, building community rather than exploiting isolation. Monetization doesn’t have to mean manipulation.

Jesse Schell’s question at GDC stayed with me: What games can only be made with AI? That’s where the real opportunity lies — not in replicating the fast food model at greater speed, but in exploring creative territory that simply wasn’t possible before. Games that remember every decision you’ve ever made. Worlds that respond to who you actually are. Experiences that feel genuinely personal, rather than algorithmically engineered to keep you scrolling.

Games Can Be More

Games can carry meaning beyond just being fun. Many educational games are designed for school-age children to learn math, chemistry, language, and more. We also see interactive exploration games in museums aimed at helping visitors learn more about the exhibits. Some games hold social significance as well.

This demonstrates that games have tremendous potential to change player behavior — for better or worse. The question is: which direction will we choose?

Of course, everyone plays games for different reasons and enjoys different genres. But when player behavior resulting from precise design becomes harmful — such as addiction or gambling — how should players protect themselves? And for developers whose original intention was to design something meaningful, how do we apply these powerful technologies while resisting commercial pressure?

The fast food model works because it’s optimized for short-term engagement. But players are becoming more discerning — they’re starting to recognize exploitative patterns and seek out alternatives that respect their time and intelligence. The market is slowly responding.

For designers, the choice is real: we can use these tools to build better dopamine traps, or we can use them to make something we’re actually proud of. The technology doesn’t decide. We do.

So here’s my answer to the question I started with: yes, AI could accelerate addiction-driven design to a scale we’ve never seen before. But it could also enable a generation of games that are more human, more meaningful, and more worth playing than anything that came before. Which future we get depends on the intentions we bring to our work — and whether we’re willing to ask harder questions than “how do we maximize retention?”

References

[1] Shaw, Caryl. “Reclaiming My Soul: Why I Quit Making F2P Social Games.” GDC Next 2014. GDC Vault. Accessed January 22, 2026. https://www.gdcvault.com/play/1021476/Reclaiming-My-Soul-Why-I

[2] EA Official Statement on Dynamic Difficulty Adjustment
https://www.ea.com/news/fair-play-and-dynamic-difficulty-adjustment

[3] NME — “EA granted patent for dynamic difficulty adjustment technology” (2021)
https://www.nme.com/news/gaming-news/ea-granted-patent-dynamic-difficulty-adjustment-technology-2915676

[4] Slate — “Electronic Arts’ FIFA, Madden, NHL Lawsuit: Dynamic Difficulty Adjustment” (2020)
https://slate.com/technology/2020/12/electronic-arts-fifa-madden-nhl-lawsuit-dynamic-difficulty-adjustment.html

[5] Zynga Engineering Blog — “Portfolio-Scale Machine Learning at Zynga” (2019)
https://medium.com/zynga-engineering/portfolio-scale-machine-learning-at-zynga-bda8e29ee561

[6] Towards Data Science — “Deep Reinforcement Learning in Production” (Zynga ML Team)
https://towardsdatascience.com/deep-reinforcement-learning-in-production-7e1e63471e2/

[7] MDPI Applied Sciences — “Predicting Player Churn of a Free-to-Play Mobile Video Game” (2022)
https://www.mdpi.com/2076-3417/12/6/2795

[8] arXiv — “On Analyzing Churn Prediction in Mobile Games” (2021)
https://arxiv.org/pdf/2104.05554

[9] ResearchGate — “Player Churn Prediction Research” (2024)
https://www.researchgate.net/publication/390168979