Teleological understanding of actions

All organisms pursue goals that promote their survival and reproduction. For many species, the small repertoire of basic drive-related goals and behavioural means to achieve them are genetically specified and are triggered by appropriate releasing stimuli. Other species, most notably, but not exclusively, mammals and birds, have much more flexible ways to accomplish their goals, tailoring their behaviour to the local environment and adjusting it according to their past experience with similar situations. Humans also display adaptability and learning capacities in their goal-directed actions, but they are flexible in another way as well: they regularly pursue goals that hold only a remote connection to their survival and reproductive aspirations (e.g., climbing Mount Everest), and in many cultures even the most basic survival needs, like consumption of food, are fulfilled by chains of goal-directed sequences composed of actions (e.g., sowing seeds, calling a restaurant to reserve a table) that are far detached from their biologically useful outcome. Beyond pursuing goals, animals may also benefit from observing other individuals' behaviours because they may provide information (1) about what the observed agents would do next and (2) about the environment. Interpreting behaviours as goal-directed actions is 2 especially valuable as it carries direct information about likely future events (the expected outcome) and its context (e.g., potential food source). What kinds of information are available in a goal-directed action that would help observers to interpret it as a means directed at a certain end (Csibra & Gergely, 2007)? First, as in many domains, statistical information about the co-occurrence of observed behaviours and their outcomes could assist the formation of bidirectional action-effect associations. Such information may contribute to gradual learning about goal-directed actions, but does not help when an observer is confronted with novel instrumental behaviours. The second source of information for interpreting actions as goal-directed comes from the observer's own motor (or more generally, instrumental) competence. One can rely on the internal mechanisms of motor planning, which chain actions to bring about desired outcomes, to simulate observed actions and to link them to their likely goals. While such simulatory mechanisms play important roles in action prediction (Csibra, 2007), their use is restricted to actions that are within the observer's own motor competence, and are unavailable for interpretation of the behaviour of unfamiliar agents or novel actions of familiar actors. We have proposed that a third source of information that, beyond temporal association and motor …