Investigating the relation between object affordances, spatial compatibility and human-robot interaction

Abstract

In their everyday life, humans interact with a multitude of objects. The ability to interact properly is made possible by their knowledge of the object properties and how they must be manipulated in order to implement actions consistent with the person's will. Gibson (1979) coined the term affordance, understood as the instrumental property of the environment that allows the individual to know how to act. In the first theorization of the term by Gibson, we find his definition: "The affordances of the environment are what it offers the animal, what it provides or furnishes, either for good or ill. The verb to afford is found in the dictionary, the noun affordance is not. I have made it up. I mean by it something that refers to both the environment and the animal in a way that no existing term does. It implies the complementarity of the animal and the environment.” (Gibson, 1979, p. 127). According to this theory, the individual is immersed in an environment that gives him information on how to act. The term affordance was then evolved and refined until the definition of the term "micro affordance" by Ellis and Tucker (2000). Micro affordances point to the implications of the motor system when perceiving an object. In fact, according to this definition there would be internal representations that mediate the perception of the object and a direct association between vision and action, where the perception of objects is not only related to the semantic knowledge of their characteristics. The association of vision and action would evoke representations of an action when you see a specific object, this activation would also depend on the intentions of the individual. The authors came to this theorization after a multitude of experimental observations where subjects were faster in recognizing an object orientation when its handle was in the same position as the response. However, there is still no unilateral agreement on the effects observed by Tucker and Ellis, and many studies tend to see spatial compatibility effects toward the functional part of an object as governed by spatial encoding mechanisms, and are therefore not necessarily related to motor components regarding the actions that are potentially achievable with the object being observed. Many studies have shown that it is not easy to dissociate spatial encoding effects from spatial compatibility effects related to the affordances of an object. Examples of criticisms to the affordance theory used to explain these phenomena stem from Cho and Proctor ‘s (2010) studies, in which the authors have shown how the compatibility effect found by Tucker and Ellis is primarily due to spatial encoding and not to the motor intentions of the subjects. In their studies they have shown how the response mode does not affect the spatial encoding effects, as a result there is no link between object affordance and motor system. The objection often raised by the affordance critics concerns the fact that the stimuli used in the experiments are usually asymmetric and, given this asymmetry, what is being observed is a spatial encoding effect. In the experiments presented in the second chapter of this work, we will try to clarify whether the affordance effects can be attributed to the activation of objects affordance or they are due to a generic spatial encoding effect. Many studies have shown how the visual perception of an object is influenced by it being seen when manipulated by another individual. The influence of the presence of another person manipulating the object is understood as a proof of the activation of a motor simulation mechanism related to actions that can be accomplished with the object. The presence of such mechanisms demonstrates that the objects are perceived according to their functional properties and are linked to motor mechanisms, so that object encoding does not only concern aspects related to spatial perception, but to our own and other people's motor intentions. In the third chapter of this work we have made three experiments to evaluate how the vision of hands approaching an object can be used to clarify whether the effects usually observed are due to the spatial encoding of the stimuli or to objects affordance. Chapter 4 lists the results of an experiment that aims at clarifying the mechanisms involving the affordance of pairs of objects and the perception of motor intentions in robotic agents. Experimental evidence has shown how the presentation context modulates the affordance effects. E.g., presenting an object together with another object, with which complementary actions can be made, makes it easier to recognize the pair of objects. Even in this case there seem to be links between the observer's visual perception and their motor system. There is a strong preference for the active object of the pair, the one with which more refined movements are performed, and that is used on the passive object. Even in the case of objects shown in pairs, the literature shows evidences that the performance of those recognition tasks is modulated by the presentation of other individuals manipulating the objects. (Laverick, Wulff, Honisch, Chua, Wing & Rotshtein,2015). It has been proposed (Gallese & Sinigaglia, 2011) that individuals are able to understand the intentions of other individuals by means of motor simulation mechanisms. During the observation of another individual acting, it is possible to understand the outcome of the actions, and therefore their intentions. This happens because we can represent internally the movements we are observing, and thus understand what will happen shortly thereafter. These motor simulation mechanisms are easily explained during the interaction between human beings, but it has to be clarified how a human being perceives and internalizes the intentions of a robotic agent. In fact, while we can easily recognize ourselves in another human being because of having similar body features, with the same biological motor constraints; it is unclear how the actions of a robot and consequently its intentions can be perceived automatically. To clarify the underlying mechanisms of a possible human-machine interaction we have investigated to what extent two robotic hands are perceived, and their influence on motor simulation mechanisms involving pairs of objects that can be used together. In our experiment we asked the subjects to classify pairs of objects semantically, by manipulating the pair composition and position of the objects to suggest a possible action for the experimental subjects. Prior to the presentation of the experimental stimuli we presented human hands and robotic hands images to evaluate the effects that these may have on the perception of the objects presented afterwards. Our hypothesis is that, if the humans show mechanisms of motor simulation with robotic hands, then it is possible that similar interactions to those typically found among human beings occur during the interaction with a robotic agent.