Marynel Vázquez, Alexander May, Wei-Hsuan Chen, Riley Harmon &
Aaron Steinfeld, 2009-2010

Can robot deception imperceptibly change user experience? What happens when users notice? Is this ethical?


We typically expect robots to operate as advertised, without hidden intentions. Our attributions toward these machines are highly influenced by paradigms concerning how good and reliable electronic systems are for certain tasks in comparison to human perception. In particular, we tend to favor robots for jobs that require memorization and keen perceptual abilities, thus making them valuable for judging objective results which require high precision. What if stereotypes regarding robot behavior were used to trick us?


We wanted to see if human expectations for robot characteristics could be subverted to alter user perception. To do this, we studied the effect of robot deception in the context of a simple, multi-player, reflex game. The robot was designed to deceive game players with the hope that they would want to play longer and with more interest, while eating healthy food.

Our robot does not participate in the game as a competitor, but is in charge of declaring who wins or loses. In this context, the robot can imperceptibly balance how much players win, due to its implied ability to perceive faster than the users. Theoretically, the balancing behavior should increase general motivation and interest in playing due to a more balanced frequency of winning.


Our multiplayer robotic game was designed with a food motif in mind. This is due to our goal of getting people together to eat healthy food. In practice, we wish players would get so engaged that they could even forget they are eating. Players should not suspect our persuasive robot of deception or dishonesty; though, they are led to believe that the only factor for determining who wins or loses is their response time.

The main objective during a round of the game is detecting the vibration of a specific “target”. The first of 4 players to react by pressing a button is the supposed winner. The fastest player gets to consume one type of food, while the rest of the gamers eat another.

The reason for playing with two distinct types of food is differentiating winners and losers. The food that the winners receive should be considered preferable by all players, while also being healthy. In particular, these foods could take the form of sweet pieces of fruit. Though the other type is also intended to be healthy, it should not taste as good as the first one. This type is what players get to eat when they lose; like steamed vegetables for example. Clearly, choosing the right types of food depends on players’ preferences.

A round of the game generally consists of the following steps:

  1. The robotic system waits for players to be ready to start the round.
  2. A target is chosen, so that players know what to look for during the round.
  3. A randomly–selected time starts and the robot tries to distract players by vibrating non–targets. During this step the target doesn’t vibrate.
  4. Distractions stop and the target vibrates.
  5. The robot listens for players that should respond to the target as quickly as possible.
  6. The round ends and either a player is the winner of the round, or everybody loses if nobody called the target within a short period of time.

A special case occurs when a player pushes their button before the target starts vibrating. In this situation the round ends, this player is declared to be the loser and he or she gets to eat the less–preferable type of food while the other players watch.

After a round ends, each player is in charge of making sure that all others are eating a portion of healthy food. The size of these portions greatly affect how long rounds take and the pace of the overall game. Considering the fluency of the game, small bite–sized portions are recommended.


The main body of the robot consists of a turntable, concealing electronic components from players.

Robot components

This turntable also holds a quasi–anthropomorphic corkscrew “robot representative” and vibrating plastic fruits on top of it. The corkscrew is the “game master” and, through non–verbal communication, orchestrates the flow of game rounds. The bottle opener basically acts as an organizer and arbitrator, primarily indicating the winner or loser(s) of given end conditions. The fruits on top of the turntable are vibrating objects that either become the target of a round or are used as distractions. They shake and their resonance becomes audible, thereby engaging sight and sound as participants wait for the target.

Players respond to the target by pushing the button on the tops of their respective controllers. In order to support the food motif, the controllers are small glass bottles with buttons also connected to the robot.


The corkscrew robot figurine is unable to see, hear, or speak. We designed it to use body language to express itself. By using the anthropomorphic shape of the corkscrew, the central element in the robot is able to suggest gaze and human–like emotions without implying that it had any more sophisticated abilities or functions.

When the bottle opener spins along with the turntable and stops in front of a player, it looks like its pointing at those players. By placing players in certain positions in front of the turntable, we made the robot be able to face the certain player and express the robotic intention to him or her.

The automated corkscrew indicates if there is a winner by waving up and down, expressing joy and happiness. When it puts both “arms” down, it expresses a more neutral or disappointed emotional state, useful for communicating defeat and indicating game losers.

waving waving waving waving waving waving waving


This work started as the final project for Principles of Human–Robot Interaction at Carnegie Mellon University. Special thanks to Prof. Illah Nourbakhsh for his guidance and encouragement, as well as to the Fall class of 2009 for all of the interesting discussions and helpful suggestions. ShakeTime! was presented in the AAAI’10 Robotics Workshop with support from AAAI. Thanks to Rafae Aziz for his help while designing user studies, and to all the participants who collaborated. Support for portions of this work was provided by the National Science Foundation under Grant No. IIS-0905148.