exploratory: how infants make sense of intentional action

[orbacht]

I really liked this paper. I thought the authors managed to relate innate abilities and experiential learning in a way that made a lot of sense to me. They took a position that was somewhere between Tomasello and Gergely that I think is intelligent and intuitive. The two main questions of the article were whether properties of motion are the sole basis for infants’ judgments of agency and whether the idea of a hard wired trigger is accurate. These were two of my main objections to Gergely’s argument so I am happy they are challenging them.
Woodward et al. show that infants attend to the object that is the goal when a human hand is reaching and the path of motion when a mechanical arm is reaching. This goes to show that infants are attributing intent to some human quality and not just to a type of motion. How does this relate to the teleological argument we read the other day? If understanding of intent could be solely teleological one would expect the infant to attend to the goal object in both cases. There must be qualities other than type of motion that indicate to infants that agents are acting with intent. Also, how does the fact that infants follow the “gaze” of fuzzy balls without eyes but that move with purpose fit into this? Would it be possible to train infants to come to understand machines as intentional agents? If it were possible to raise a child in an environment where robots were everywhere from the day they were born would they understand the machines as having intent?
Woodward et al. try to zoom in on what features are triggering this attribution of intention in the study with the gloved hand. I was really glad that they ran this study because I was concerned that infants were cluing into the fact that machines move in a different fashion to humans when making attributions of intention. I was wondering if infants would become confused if there was a really human like mechanical arm doing the reaching. This study showed that infants attend to the reach and not the grasp when a human hand is wearing a mechanical glove. This shows that infants are using some textural quality such as skin to identify whether the agent is human and therefore intentional. The finding that when infants saw the adult wearing the glove and then just a zoom in of the glove they construed the hand as an intentional agent is particularly interesting. This means that infants are using multiple features as triggers of intentionality. This finding means that the infant system for understanding intention must be far more complicated than the teleological proposal. Infants are integrating a number of factors in order to explain the motion of an agent. This finding reminds me of Lauren and my point on Monday that the built in interest in things like skin and faces would cause infants to attend to things like arms reaching more than metal claws reaching and therefore to associate those reaches with intention.
The final study mentioned in the article where demonstrators performed various actions in order to get the goal in the boxes was of particular interest to me. The fact that infants seemed to understand the reason the adults performed all of these goals and see all of these actions as a means to an end (getting the toy) proves that infants have a true understanding of causality. If infants are able to understand actions in relation to the goal it means they are able to understand that other actions could be performed instead of the ones that they just observed in order to reach the goal. This in turn means that infants can understand that goals exist in other people’s minds and that they are using their own reasoning in order to reach them, which is the beginning of theory of mind! I wonder what would happen if a non human agent was doing the acting in this study. What if it were either something mechanical, an animal, or something computerized like an arrow. This account gives a lot of weight to experience and would leave a lot of room for learning effect studies. I would be very interested to see if infants could be trained to perform differently in any of the above studies or in any similar studies. If they could it would give more support to an association hypothesis. It is possible that all of these hypothesis are working together as independent systems (again!) and that the information is being synthesized at a later point in the game.