Thinking before Acting: Intentions, Logic, Rational Choice
Olivier Roy

Abstract:
In this thesis I propose a theory of decision making that does justice
to the idea that human agents can form future-directed intentions, but
which at the same time capitalizes on the resources of contemporary
theories of instrumental rationality and dynamic epistemic logic. The
result is a more all-encompassing picture of practical reasoning for
planning agents. I show that such a broad approach genuinely enriches
existing models of rational decision making, as well as the
philosophical theory of intentions.

In Chapter 2 I show that the introduction of future-directed
intentions does indeed broaden the explanatory scope of
decision-theoretic models. The volitive commitment of
future-directed intentions allows one to go beyond traditional
decision-theoretic reasoning by ``breaking ties'' between equally
desirable options, and thus provides a straightforward anchor for
personal coordination.

In Chapter 3 I consider coordination, mostly in ``Hi-Lo'' games. I
show that intentions do indeed anchor coordination in these games, in
a way that naturally generalizes their ``tie-breaking'' effect in
single agent contexts. At the end of the chapter I look at how
intentions can anchor coordination in the general case. This allows
to revisit important claims in the planning theory concerning
``shared agency'', and in particular to circumscribe better the extent
of this phenomenon.

In Chapter 4 I turn to two facets of the reasoning-centered commitment
of intentions, namely the filtering of options and the focus on
means. I show that they can be studied by means of two simple
operations which transform decision- and game-theoretic models. In
contexts of strategic interaction, these operations acquire an
important social character, that has not yet been studied in
philosophy of action.

In Chapter 5 I use dynamic epistemic logic to bring
the considerations of the previous chapters under a single umbrella. I
show that an important aspect of the volitive commitment used to
account for coordination with intentions has an echo in the filtering
of options that I define in Chapter~\ref{cha:chapter-4}. This
observation triggers a natural generalization of the idea of
filtering, which takes into account the information that agents have
about their own intentions and the intentions of others. By the end of
the chapter I explore two other issues at the intersection of planning
agency and instrumental rationality, namely the condition under which
intention-based transformations of decision problems foster
coordination and become ``enabled'' by the elimination of dominated
strategies.

In Chapter 6 I look back at this theory from a philosophical point of
view, and investigate the question of how the norms of consistency and
coherence which apply to intentions can be explained. In contemporary
philosophy of action there are two main takes on this issue, called
the ``cognitivist'' and ``agency'' approaches. Here I explore an
alternative one, \emph{hybrid pragmatism}, which stands half-way
between cognitivism and the agency approach. It is based on the notion
of ``acceptance in deliberation'', a cognitive state which has so far
attracted little attention. I argue that hybrid pragmatism is a
plausible alternative to the two main contemporary approaches, and
that its use of acceptances provides a more complete picture of how
future-directed intentions make their way into practical
reasoning.


Keywords: