33

I've recently seen the term skinner box thrown around more often as a derogatory term by reviewers for free-to-play games that use predatory practices to encourage players to spend more money than they actually intend to.

The term seems to originate from the research of a well-known biologist, B.F. Skinner, referring to a device that is used to train animals in biological experiments (such as training mice to solve mazes, or training a parrot to associate a colour with a button). See https://psycnet.apa.org/record/1998-06434-011 for an overview or https://www.sciencedirect.com/science/article/pii/S0003347270800343 for an example. Or read the original paper here. The animal learns to respond in a certain way to a stimulus and receives a food reward for correctly doing so in a reinforcement learning process.

Some semantic change seems to have happened when this word is now used in the game (review) context. Games don't literally trap people in tiny cages and force them to do certain actions to stay alive (that would be a rather unethical experiment in many ways), so it clearly means something different here.

There are many techniques modern game designers employ in free-to-play (and recently, also in paid) games to encourage users to spend money, from the straightforward to the subtle. Some examples are (with an example game that includes this technique):

  • Allowing the purchase of actual straight up power increasing bonuses (skipping content or in multiplayer having advantages over those who don't), Example game: Maplestory.
  • Purchasing 'cosmetics', making the characters look aesthetically better in some way for those who spend. Example game: Dota-2
  • Unlocking 'more stuff' I.e. the gacha or lootbox system where spending money unlocks additional gameplay options via more available characters or items, but what you get is random. Example game: Genshin Impact.
  • Skipping some kind of progression system by spending money. Example: Bloons TD 6
  • Gating 'game time' in some way, requiring a resource necessary to continue playing and then making more of it available for real money. Example: Candy Crush.
  • Making things happen faster in simulation and strategy games when you spend. Example: Clash of Clans.

It's not entirely clear to me which of these techniques fall under the term, i.e. "which are skinner boxes, and which are not?",or if reviewers are using it more as a blanket derogatory term for those types of transactions they personally emotionally disapprove of. Is there an objective or reasoned definition of which variants/strategies generate the 'conditioned responses' Skinner has written about in his papers?

  • What does it mean for a game to be a 'skinner box'?
  • Specifically: What are the requirements?
  • Is there a generally agreed-upon definition?
  • Could it be that this varies from person-to-person?

Footnote:

How it could be subjective (conjecture): what causes addiction, which is very similar to a conditioned response in some ways, varies from person-to-person, thus which techniques qualify as sufficient conditions for a game to be included in the set of 'skinner box' titles are subjective if this is the correct definition for 'skinner box element':

a to the person reviewing it addictive or enticing element that involves spending real money and conditions them to spend more

3
  • 1
    Reading followup publications, it seems people experimented used a random interval (e.g. the number of presses needed for the next pellet is from 2 to 8), while modern games use just a random probability (1 in 5) which could need more than 8 presses, but both qualify as 'variable ratio/skinnerbox'. I suppose the specific statistical distribution could improve the addictiveness, as I see many of these types of mechanics have complex 'guarantees' against bad luck of some sort (e.g. you get a 3% super-rare guaranteed every 100 rolls). Popular games do data science to achieve maximum profit.
    – aphid
    May 7, 2021 at 9:08
  • Modern games often use far more sophisticated "RNG" algorithms than just random probability. There is a saying among game designers: "The generation of random numbers is too important to be left to chance". Fudging RNG results in order to provide the best possible (or most profitable) game experience is a popular tool in the toolkit of every experienced game designer. "number of kills needed for the next loot drop is from 2 to 8" isn't a particularly unusual mechanic nowadays.
    – Philipp
    May 7, 2021 at 12:00
  • 2
    I don't understand why people would play those lousy games when there are much better games. May 7, 2021 at 16:42

3 Answers 3

82

The term 'Skinner Box' in game design discussions refers to one of the key findings of the experiments. Skinner found that the behaviour of the animal could be changed based on how the experiment trained (conditioned) them. There were 2 cases that form the basis for the term as it's used today:

  1. The animal had access to a button, and when they pushed the button, a food reward was given every time. In these cases, the animal would push the button a few times, and then stop when they no longer wanted food.

  2. The animal had access to a button, and when they pushed the button, a food reward was given with a random chance. In these cases, the animal would repeatedly and obsessively push the button, well beyond the point at which they had a desire for more food.

In games, a Skinner Box is any mechanic that uses random chance to increase engagement or spending of players. If you've ever spent hours and hours grinding for a chance at a rare loot drop, if you've ever spent more money than you wanted on loot boxes hoping for a specific reward, or for that matter, spent too much at a casino, then you've fallen into the engagement trap of a Skinner Box.

Use of random chance in games isn't inherently bad, but when it's used to drive player engagement or spending, it can end up being psychologically manipulative or even abusive. There are significant concerns about the long term effects of these incentive systems, especially in children.

So of the examples you listed, only the third one, the 'gacha or lootbox system' is technically a Skinner Box, but the others are examples of different mechanisms designed to drive spending or engagement.

1
29

You're overthinking this. A "Skinner Box" is not a genre of game with a well-defined scope. It's just a derogatory metaphor.

In a real-world Skinner box, an animal is rewarded for performing a simple task. By comparing that to a game, you're saying the game is not enjoyable for its story or challenge, but rather for its stupidly simple task-reward cycle. However, depending on what you consider "stupid" and "simple", that could be any game.

3
  • 27
    "that could be any game". Or you know, any Q&A website with a reputation system. :) May 7, 2021 at 6:31
  • @EricDuminil If the goal becomes collecting points to the active detriment of the main goal, then probably yes
    – mishan
    May 7, 2021 at 8:13
  • 2
    Or even if the game is fun and well-written, you can get to a point where your motivation for playing it is basically the Skinner-box reward cycle. e.g. for a while I was playing a lot of Diablo II (single player, not online) to see how much loot I could collect. Once I read about the concept of a Skinner Box, I realized that's how I was playing D2, not so much for the challenge of actually beating the fights. Posting on Stack Overflow has some of the same reward cycle, but is also actually useful in contributing to a repository of knowledge, so I keep doing it. May 7, 2021 at 10:34
17

While there is probably some wiggle-room as to what does or doesn't fall under the definition, a Skinner box classically will have two things:

  • A compelling "reward"
  • An interval between when "rewards" are awarded

The simplest version of this is the "loot box"; consider a game like Overwatch -- whenever there's new content added, you can't buy it directly. Instead, you need to earn the latest version of lootboxes by playing the game (or by spending real money).

It turns out, the delay between "doing the thing" and "receiving the reward you want" is psychologically very insidious. Having a random chance to achieve the desired result (instead of a guaranteed thing) increases player engagement due to sunk-cost and gambler's fallacies.

The more general term for this is "operant conditioning", and my guess is that some reviewers are conflating the two terms, as many of the examples you list are examples of operant conditioning without a direct "Skinner Box" analog.

5
  • 4
    There was an episode of the YouTube show "Mind Field" where they put a person in a literal box with a big red button that does absolutely nothing, and a slot which spits out cash at a fixed regular interval, and gave the person an hour with no explanation whatsoever, just to see what behaviors they would develop, thinking it in some way was connected to the money coming out. If your video game feels like this, it might just be a Skinner box. May 6, 2021 at 19:30
  • I believe the variable interval or variable ratio is the key here, and in Skinner's experiment, those were the situations that were most stress-inducing and created the most response.
    – A N
    May 6, 2021 at 19:46
  • 3
    @AN Indeed. Fixed intervals makes it very easy to identify when the rewards have stopped. But once you add that tiny bit of randomness into it, people willfully persist in the behavior for longer. May 6, 2021 at 20:08
  • I think in the case of the episode I linked, they kept it fixed for simplicity, but the "contestants" were only in the box for an hour, apparently not long enough for any of them to pick up on the fixed interval. If your gameplay is not time-limited in that way, you would probably benefit from some randomness. May 7, 2021 at 14:14
  • The important part of a Skinner Box is that there is a behaviour developed as a response to the reward but the behaviour does not affect the frequency of the reward. In other words, the subject of a Skinner Box would develop a superstition concerning the reward model. If that element is not present then it probably is only an operant conditioning.
    – Simon
    May 8, 2021 at 16:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .