Software for training dogs and cats to use electronic interfaces
The CleverPet curriculum is a series of challenges designed to a get a dog, cat, or really any similarly sized animal all the way from wariness to mastery of a simple electronic interface. Developed by CleverPet’s founding team of neuroscience PhDs, it is the product of decades of combined experience training animals with software. This entire training curriculum is now publicly available to customize or modify.
Tweaking one of the challenges below so it works well for the player you’re training can help accelerate the process and make it more fun for all involved. That noted, each of the challenges is designed to act as a gate for subsequent ones: skipping ahead is not recommended, and will usually lead to a slower, or even entirely ineffective, training process.
We’ve had to get precise with some of our words.
Player: Any dog, cat, person, or other animal who is playing with a Hub.
Foodtreat: A food reward. E.g., a dog treat, cat treat, or piece of kibble.
Report: A single row of data describing everything that a player did during an interaction.
Interaction: A presentation of lights, sounds from a Hub, and the corresponding responses of a player, ending with a report. Nearly always, an interaction begins with the Hub doing some things, the player doing some things in response, and then the player getting some feedback as to whether they did the right thing.
Challenge: A series of one or more interactions, usually of progressively increasing difficulty, and often designed to teach the player a particular skill.
Level: A stage of difficulty within a given challenge. Lower levels are easier, and each challenge usually has a fixed number of them.
Game: A fuzzy term, currently without precise technical definition, sometimes used interchangeably with “challenge”, but which may consist of multiple challenges.
Computer-supported training requires recording how well a player is doing and adjusting the game accordingly. This can often take weeks or months.
To get your player fully trained up with the Hub, start with the standard CleverPet curriculum (the challenges beginning with 0). If your player is new to CleverPet you should start them off with challenge 000 and only shift them to the next challenge when you’re certain they’re ready. If you shift them up too early this can significantly increase the time required for training, since the player will at first learn the “wrong” thing and then need to unlearn it afterward. It’s much better to have your player learn a little too slowly than a little too quickly.
Roughly speaking, the learning steps required for getting your player to productively using the Hub are, in order:
The goal of this challenge is to help your player get comfortable with the Hub’s sounds and movements.
The Hub’s dish will rotate out and offer a free foodtreat to your player at varying intervals. If your player does not take the foodtreat, the dish will stay out for progressively longer periods of time.
Challenge logic: This challenge has 6 levels with 6 corresponding foodtreat offer durations. A foodtreat is automatically offered to the player for the corresponding time. If the foodtreat isn’t eaten the level is increased and the offer duration will be extended. If the foodtreat was eaten the level will go down and presentation time will be shortened. If 3 foodtreats were eaten in the previous 5 interactions, the challenge is completed and the performance array will be reset.
A foodtreat is still offered periodically “for free”, but the player will also earn a reward when they press a touchpad — even if just “by accident”.
This challenge is designed to have players begin to associate their own behavior with getting a foodtreat. Through a combination of classical and operant conditioning, players begin to understand that the Hub is a tool they can control to get tasty snacks.
Challenge logic: The challenge consists of 4 levels with accompanying time-out speeds and foodtreat presentation intervals. If 3 foodtreats were eaten in the last 6 interactions, the player levels up. If 4 foodtreats were left uneaten within the last 6 interactions, the player levels down. When the player reaches level 4 and eats 3 foodtreats, the level resets back to level 4.
This challenge takes off the training wheels and no longer offers your player free foodtreats. Now your player will need to press a touchpad to earn a reward.
Through repeated action and feedback from the Hub, players learn that the front of the Hub — where the touchpads are — is the most interesting part. Before this challenge, a very patient player could hang out and simply wait for foodtreats to arrive on their own.
Challenge logic: This challenge uses a combination of performance and a timer to progress the game-play. There are 3 levels and 3 corresponding timer durations (10 minutes, 10 minutes, and 5 minutes respectively). The timer starts running, and the player can play so long as the timer hasn’t expired. During the challenge the player levels up if 10 foodtreats have been eaten. When the timer expires, the level is checked and the corresponding timer length is loaded into the timer. The current count of foodtreats eaten is cleared and new interactions can continue to be played while the timer is running. Once the player reaches level 3 and eats 10 foodtreats the max level is reached. Leveling down requires many misses, since the player would need to fail to consume 99 foodtreats in one level. When this happens, the timer is immediately reset and the duration is adjusted. Another way to level down is be to ignore the touchpads, which will make the current interaction time-out.
In this challenge only two of the three touchpads are lit, and your player must learn that only pressing illuminated touchpads will result in a reward. This is the first time the Hub’s lights begin to serve as meaningful cues. Using trial and error, your player will gather evidence and figure out how the Hub works, and what to do to earn a reward. All attempts to get kibble (whether successful or not) are part of the learning process. As the player learns, they will hit the unlit touchpads less and less often.
Challenge logic: This challenge has 2 levels. The first level chooses 2 targets at random. If 1 of the targets is pressed, the accurate flag is set and a foodtreat is dispensed. If accurate, a positive interaction is added to the performance history. When 18 of the previous 20 interactions are accurate, the player levels up. At level 2 the interaction is identical save that if the player has gotten it wrong they will have to redo the interaction: if the player pushes the wrong touchpad, the player will be presented with the same targets on the next interaction. This prevents the player from developing and keeping a favorite touchpad. If the interaction times out, a miss wil be added to the performance array. Between interactions there is a random wait of between 1 and 8 seconds.
Only one touchpad is lit. This is the challenge where most players will start to “see the lights”. If your player wasn’t using the information provided by the lights on the previous challenge, they could expect to get rewarded about two-thirds of the time, just by chance. Now, chance is only 1 in 3 guesses. The lights are twice as important as before!
Challenge logic: This challenge has only 1 level. if 30 successful interactions were played in the last 50 games, the challenge is done (will reset). Every interaction, 1 random random target will be picked and compared with the pressed touchpad. The player’s performance is recorded, and only correct presses count. When an incorrect touchpad is pressed, the next interaction will be a redo interaction with the same target. There is a random between interactions pause ranging from 1 to 8 seconds
This challenge continues with only one lit touchpad, and now your player will need to engage even more consistently and accurately. There are some sneaky ways for players to get enough rewards without always paying attention to the information the lights provide. This challenge improves on the progress your player has made by making such strategies much less rewarding.
Challenge logic: This challenge has only 1 level. If 40 successful interactions were played over the previous 50, the challenge is complete (will reset). Every interaction, 1 random target will be picked and compared with the pressed touchpad. Performance is tracked, but time outs have no effect on the game’s estimate of the player’s performance. When a the touchpad is pushed, the next interaction will be a redo with the same target. There is a random between-interaction pause ranging from 1 and 8 seconds.
In this challenge, immediately after your player presses one illuminated touchpad, a different one lights up and must be touched. Your player now has to press two touchpads in a row to solve one puzzle. Chaining behaviors together requires that your player complete multiple correct actions in sequence to earn a reward. Your player must learn to stay attentive for a second instruction even after the first press.
Challenge logic: This challenge has 4 levels. With each higher level the maximum reaction time decreases. One interaction consists of 2 moves, in each move 1 touchpad lights up and has to be touched to advance. The reaction time is the time between the first and second moves. If 17 successful interactions occur over the previous 20, the player levels up. After 17 misses the player levels down. On a miss there will be a between-interaction delay. For the player’s first move multiple touchpads can be touched, as long as the lit up touchpad was one of them (player has unlimited opportunities to try and get it right). The second touch, however, allows only the lit up touchpad to be touched, and the interaction will result in a miss if any of the lit touchpads is pressed more than once, or if the unlit touchpad is pressed at all.
After the first press, the other touchpads light up and your player will need to choose the brightest one. The touchpads start out with very different brightness levels and gradually become more difficult to tell apart.
Your player now has to make a more careful comparison between the target (brighter) and distractor (dimmer), and will gradually improve as they learn to attend to the relevant feature. This skill will come in handy during the next few challenges.
Challenge logic: This challenge has 4 levels with 4 corresponding minimum and maximum levels for distractor touchpad intensity. At the start of an interaction, the touchpads are randomly shuffled and a distractor intensity level is picked randomly. At the 4th level random probe interactions are introduced with randomly chosen higher (i.e., more challenging) distractor intensities.
The first target in the interaction is a single target which the player gets unlimited retries to try and touch. The second target is a different pad from the first target and a distractor is added to the interaction, which is also different from the first target. If the player misses, the same targets will be presented in the next (redo) interaction, except on probe interactions.
Performance on redo interactions and intensities outside the threshold interval aren’t counted. There is a between-interaction wait time of 6 seconds when the player misses. When a player has 40 successful interactions within the previous 50, the player will level up. The player cannot level down.
This challenge engages your player with sequences of actions. For a successful interaction, your player must press the brightest touchpad accurately twice in a row.
During this challenge, the Hub gives your player information that will help them plan their next move. The dim touchpad during the first sequence becomes the the player’s next target. Clever players will catch onto this and use this “hint” to prepare their next press.
Challenge logic: This challenge has only one level. The challenge has 3 moves per interaction: 1 stimulator round and 2 moves that are part of a sequence. The stimulator round lights up all 3 touchpads and waits for the player to touch any of them in any combination.
For the player’s second move the touchpads light up 1 random target and also light up the next target at a lower intensity. The target, and only the target pad, needs to be touched for the interaction to continue to the next round in the sequence.
When the player has had 40 successful games in the last 50 games, the player has reached the maximum level.
The sequences get longer! Now, your player is challenged to complete patterns of up to nine moves in a row. Beginning with move sequences of length three, the sequences get progressively longer with the player’s success, and shorter when the difficulty is too high.
To master this challenge, players need to stay attentive and accurate over progressively longer periods of time. They are given a few “lives” so that one missed touchpad doesn’t always send them back to the beginning.
Challenge logic: The sequence increases if 12 of the last 15 interactions were successes, it also decreases if 12 of the last 15 interactions were misses.
The first interaction in the challenge is a “stimulator round” where all 3 touchpads light up and any combination of pads can be touched to advance the interaction. The second interaction starts the sequence. At every step of the sequence the current target will light up, along with following target at lower intensity. The target must to be touched exactly, with no combinations or double touches allowed. On a wrong touchpad touch the player loses a “life” and is allowed to continue the interaction. If all 3 of the player’s lives run out the interaction becomes a miss.
For the first time, the Hub introduces colors (blue and yellow - yes, dogs and cats can see color!) as a meaningful signal. Each time a touchpad is pressed, its color will change. Your player’s job is to make all the touchpads match within a limited number of overall presses.
Up until now, there was always one correct answer about which pad to touch. This is the first challenge where each puzzle has more than one correct solution, and it showcases your player’s abilities solve problems creatively and in their own way.
Challenge logic: This challenge has 4 levels. Every level has a corresponding maximum number of touches. The challenge uses two different colors, yellow and blue.
At the start of an interaction every pad randomly gets one of the 2 colors. Every touch, the color of the touched touchpad advances to the next color. This all happens in a while loop, and with every iteration the state of the lights gets checked for a match of all 3 pads. If a match is found the while loop is exited and the player gets a foodtreat. Every touch decreases the maximum touches counter, if the touches are about to run out, the ‘Do’ sound will be played. When the remaining touches become zero, the interaction becomes a miss. On a miss the player’s next interaction will be a retry interaction with the same starting state.
If 4 successful games are played within the last 5 games the player will level up. On 3 missed games in the last 5 games, the player will level down.
This challenge is similar to Matching Two Colors, but now the pads can also turn white.
By adding another color, the number of possible solutions grows exponentially! Adding a third color means there are over 24 initial combinations to solve, and many different paths to solving a given puzzle.
Challenge logic: This interaction has 4 different levels with 4 corresponding touches. There are 3 different colors: yellow, white and blue.
At the start of the interaction a random combination of colors is chosen for the touchpads (starting state). Every iteration of the main while loop checks if the colors match and advances the color of the touched touchpad to the next one. If the player is about to run out of remaining touches, the ‘Do’ sound is played with every touch as they are close to running out.
When all touches are used, the interaction becomes a miss, if a match is found, the player gets presented with a foodtreat.
If 4 successful games are played within the last 5 games the player will level up. On 3 missed games in the last 5 games, the player will level down.