A Brief Introduction to Affordances

Post date: Apr 12, 2017

What exactly are affordances? From the word, we can see the root word being 'afford', which describes the instance where something or someone provides or supplies an opportunity or facility.

So within the context of my research, what does this mean?

The term 'affordance' was coined by psychologist Dr. James J. Gibson in his notable work The Theory of Affordances which he published in 1977 [1]. In his work, he described the relationship between objects and the actions which they afford, relating to what actions these objects innately provide for us to do with them. There has been an extensive number of studies [2-6] which followed which suggest that our brains relate objects to actions, as our brains fire signals when we perceive objects and any sort of manipulation of said objects. In another way, we can also imagine a variety of actions just by simply observing an object or anything in our environment.

Even if we do not see the object, just simply observing the way the hands move or how we perform the action is enough of an indicator to what objects we may be using [7]! This fascinating study by H. B. Helbig shows the power of the human brain in inferring the objects that can be used in specific activities. To summarize this study, a set of volunteers were asked to watch videos where objects were hidden from a certain action demonstration, but the hands of the demonstrator were kept visible to the viewer. The volunteers were asked to indicate what object they think is being used in the videos from a list of objects which were things that were not in the video. However, there was always one item in that set which was similar to the one used in the video and that was the indicator that affordances play a key role in identifying objects! They did conclude and find that volunteers can relate certain motions to certain items.

(Source: http://www.doctordisruption.com/wp-content/uploads/2013/09/switches.png)

For example, we have come to associate buttons with the "pushing" action. Even though we may find buttons of all shapes, colours, and sizes, we already know in our minds that we can push them to make them work. Observing a knife already primes us in thinking about the "cutting" action because we can see how sharp it is and how it is usually used and manipulated by us humans. Just as in the picture above, there are a lot of examples of where these affordances are in our daily lives! It also explains why we may be able to relate certain actions to items we have never seen before just by associating features and actions we know about or those we have been previously exposed to.

As a result of these studies, researchers in robotics have been applying the theory of affordances in robot learning. If a robot is able to identify objects in its environment, then using affordances, we can already know what kind of actions we can perform with the item. This would however require that a knowledge system or model is developed which can be used as reference in manipulations. Our group has been working on a means of representing knowledge known as FOON which graphically describes activities in terms of the objects and the motion or action taking place, while noting the effects of these actions on these objects (if any). Many researchers have taken the approach of representing these effects through other means such as probabilistic models (Bayesian Networks, Markov Models, Conditional Random Fields, etc.) which pretty much learn about the most likely occurring association of objects, actions, and effects.

It is my hope to learn more about how these models work and to eventually come upon a means of building such models and executing meaningful robot manipulations with them.

References:

[1] - J.J. Gibson. The theory of affordances. In R. Shaw and J. Bransford, editors, Perceiving, Acting and Knowing. Hillsdale, NJ: Erlbaum, 1977.

[2] G. Rizzolatti and L. Craighero. The mirror neuron system. Ann. Rev. Neurosci., 27:169–192, 2004.

[3] G. Rizzolatti and Craighero L. Mirror neuron: A neurological approach to empathy. In Jean-Pierre Changeux, Antonio R. Damasio, Wolf Singer, and Yves Christen, editors, Neurobiology of Human Values. Springer, Berlin and Heidelberg, 2005.

[4] E. Oztop, M. Kawato, and M. Arbib. Mirror neurons and imitation: a computationally guided review. Epub Neural Networks, 19:254–271,2006.

[5] G. Di Pellegrino, L. Fadiga, L. Fogassi, V. Gallese, and G. Rizzolatti. Understanding motor events: A neurophysiological study. Exp Brain Res, 91:176–80, 1992.

[6] V. Gallese, L. Fogassi, L. Fadiga, and G. Rizzolatti. Action representation

and the inferior parietal lobule. In W. Prinz and B. Hommel, editors, Attention and Performance XIX. Common mechanisms in perception and action. Oxford University Press, Oxford, 2002.

[7] HB Helbig, J. Steinwender, M. Graf, and M. Kiefer. Action observation can prime visual object recognition. Exp Brain Res 200(3-4): 251–258, 2010.