What is it to be skilled at mindreading?

• Mindreading is a joint activity you can do alone: we generate and select explanations of thoughts and action that other people would recognize as reasonable.
• Social experiences plus context tune what feels plausible.
• This explains individual differences and offers new ways to measure mindreading skill.

When we act together, is there just one way our intentions can be shared?

• Three theories: plural (one intention with many subjects), aggregate (a group agent with its own agenda), and reductive (a structure built from individual intentions).
• Each fits different phenomena.
• Living out a theory of shared intention can make it true of us.

How could we know if a machine really has a mind? Some obstacles:

• We lack a shared test, or even a shared theory, of what minds are.
• Everyday mindreading doesn’t work beyond familiar cases
• Lab tasks assume a mind to start with.

Does mindreading adhere to ‘seeing is believing’?

• In automatic responses, people behaved as if an agent had a belief even when the agent never saw the event.
• Processes behind automatic mindreading may clash with philosophical principles about mindreading.

What distinguishes genuinely acting together from acting in parallel but merely individually?

• The standard view: shared intentions make the difference.
• Our alternative: a collective goal is enough, at least on standard criteria.
• This view is mechanistically neutral: it says nothing about what’s in your head.
• Payoff: we can identify targets of explanation without presupposing claims about which mechanisms explain them.

Is commitment essential to joint action? We tested intuitions about this using controlled scenarios. Results hint at links between commitment and joint action, but do not support the view that commitment is essential to joint action.

If acting together helps children learn about minds, what kind of joint action makes that possible?

• The kind that involves shared intention demands the very mindreading skills we’re trying to explain.
• A simpler kind: joint action with shared goals, where each expects the other’s goal-directed actions to result in a common effect.
• There are at least two kinds of joint action: planning-based (shared intentions) and minimal (shared goals).

Do you need the idea of truth to understand other people’s beliefs?

• Two kinds of awareness: mindreading (tracking what people believe and how it guides action) and interpreting (using beliefs to give reasons).
• Mindreading can handle conflicting beliefs without invoking truth or falsity.
• Interpreting requires a grasp of truth/falsehood to explain why beliefs justify actions.

When we act with a purpose, are we always acting for reasons?

• Evidence from animal learning + our own self-defeating choices shows some kinds of purposive action are information-based but unreflective;
• these are unlike actions which occur because the agent takes the action to be reasonable.
• The two kinds of action require different kinds of rationalising explanation.