CritBot: The Art Critic Robot

Group project for Cognitive Science 211: Perception & Action, Professor Ken Livingston


“What is art?” is an age-old question studied and interpreted by many diciplines.  We thought: maybe a robot could finally tell us the answer. For this project class project we had to design an autonomous robot with a novel behavior.


It was my idea to embody an art critic into our small robotic agent. My goal was to have it not only move around a gallery space, but also physically demonstrate actions as interpretations of artwork. I also contributed largely to the programming of CritBot, writing code in C (Fun fact: I had no coding experience prior and had to pick it up for this project!).


  • KISS Institute for Practical Robots (KIPR) Link Controller
  • KISS Platform running C


CritBot utilized a subsumption architecture organized in states. This overall action-selection mechanism translated well to the computer programming, because we could divide the target behavior into different smaller states such as: search, escape, and evaluate.  States like “evaluate” could further be divided into subcategories of different actions.

CritBot was specifically programmed to see blobs of colors in an artwork.  When an artwork met a certain blob criteria, it would react accordingly. If CritBot likes a piece it will perform a sort of "happy dance", turning side to side.  If it dislikes something, it will make a sharp 180-degree turn and speed away. If it is unsure, it does what you may see many humans do in museums: move closer and further from the art work in hopes that changing the proximity can help one analyze it further.

Judging by CritBot's reaction to Kate's red pants in the video, I wonder what it would think of 'Untitled (Red, Orange)' by Mark Rothko, 1961.

Judging by CritBot's reaction to Kate's red pants in the video, I wonder what it would think of 'Untitled (Red, Orange)' by Mark Rothko, 1961.


  1. CritBot’s sensors allowed it have a sense of vision similar to that of a human. Its camera allowed it to recognize color, helping to indicate what was the gallery wall versus what was art, and its infrared sensors marked proximity to its surroundings. In combination, CritBot could “see” and recognize objects.
  2. There was a wonderful sense of the unexpected with CritBot.  Variables such as what angle the robot approached the wall or how much light was shining on the art, CritBot would see things differently.  The same painting could yield 1 blob of yellow reaction in its first evaluation and 7 in its second.  We may have programmed a red-heavy painting to garner a reaction of “like” only to have CritBot interpret the color as something else entirely and dislike the work completely. This gave CritBot a unique sense of autonomy.


CritBot was very successful in performing distinct tasks: Moving around an art gallery or museum space, identifying art versus the wall, approaching art then stopping in front of it, evaluating the work, reacting to it, moving onto other piece, and repeating the process. It was a unique behavior that could be performed in real-time in response to stimuli in a human environment. 


As they say, “beauty is in the eye of the beholder”.  Art is subjective, its interpretation idiosyncratic. Our first reaction to art is often one of like or dislike without much explanation for these opinions.  Thus, CritBot serves as a commentary on human nature and both the serious and comical aspects of our often arbitrary interpretations of art.  Further study could delve deeper into this question: Why do we react to art the way we do?