Partial reinforcement

Use a partial reinforcement schedule, you’ll keep people playing for longer.


  • Facebook users checking to see how many likes they received for a post and Twitter users counting followers are working on a variable ratio partial reinforcement schedule.
  • People hunting for bargains in Amazon’s gold box are working on a fixed interval partial reinforcement schedule.
  • John Hopson at Gamasutra wrote a great article on game design seen through an operant conditioning lens.


In dog training, you start by rewarding everything that looks like the behavior you want with a treat. Subsequently you only reward the exact behavior you want. Once the behavior is set, you can move to a partial reinforcement schedule where the dog is only given a treat occasionally. The dog will continue to perform the exact behavior even though it is not rewarded every time.

When the rewards are backed off, they can either be given once every certain number of times the behavior is shown (a fixed ratio schedule), or on average once within a certain number of times the behavior is shown (a variable ratio schedule). Rewarding in a specific timeframe, again either fixed interval or variable interval, can do the same although this does not provide such a strong reinforcement.

How to design for a partial reinforcement schedule

  • Keep people engaged by creating multiple goals that can be worked on simultaneously.
    • Some should be longer term fixed interval/ratio events, which happen after a certain time has elapsed/number of items have been collected.
    • Others should be shorter term variable interval/ratio events, which happen within a certain time range or number of enemies have been killed, but with an element of uncertainty built in so it’s not always clear exactly when the reward will be given.
  • Prevent people from ignoring you by using negative reinforcement. Design the system so that bad things happen if the user doesn’t visit regularly. Using a variable interval until the bad things occur will mean they don’t know exactly when to return, increasing anxiety and likelihood of checking in.