A Framework for Resolving Motivational
Conflict via Attractor Dynamics
Alejandro Jimenez-Rodriguez
(B) , Tony J. Prescott , Robert Schmidt ,
and Stuart Wilson
The University of Sheffield, Sheffield, UK
{a.jimenez-rodriguez,t.j.prescott,
robert.schmidt,s.p.wilson}@sheffield.ac.uk
Abstract. Motivation modulates behaviour depending upon contextual
and internal cues. Like animals, successful artificial agents must implement different behavioural strategies in order to satisfy dynamical needs.
Such causal factors emerge from internal homeostatic or allostatic processes, as well as from external stimuli or threats. However, when two or
more needs coalesce, a situation of motivational conflict ensues. In this
work we present a four-stage dynamical framework for the resolution of
motivational conflict based upon principles from dynamical systems and
statistical mechanics. As a central mechanism for the resolution of conflict we propose the use of potentials with multiple wells or minima. This
model leads to behavioural switching either by means of a bifurcation
or by the stochastic escape from one of the wells. We present analytical
and simulation results that reproduce known motivational conflict phenomena observed in the study of animal behaviour, in the case of two
conflicting motivations.
Keywords: Motivation · Motivational conflict · Dynamical systems ·
Behavioural switching · Attractor dynamics
1 Introduction
Dealing with motivational conflict is an important aspect of animal behaviour
[19]. Consider, for example, the conflicting need to eat or drink as studied in
[15,16], or the conflict between aggression and mating as exemplified by the
response of male stickleback fish to territory invasion by a female conspecific [1].
The former conflict is driven by opposing intrinsic needs or deficits, the latter is
driven by extrinsic factors, and in both cases conflict resolution requires multiple
needs to be satisfied.
Different types of conflict yield different behavioural patterns. Conflict may
be resolved via time budgeting, modulated by biological rhythms that occur on
different time scales [2]. Or it may be resolved by switching between alternative
behaviour systems that may be inhibited or disinhibited by the current behaviour.
New patterns of behaviour may emerge (ambivalence), and these may be related
(redirection) or unrelated (displacement) to the causes of either motivation [2].
c
Springer Nature Switzerland AG 2020
V. Vouloutsi et al. (Eds.): Living Machines 2020, LNAI 12413, pp. 192–203, 2020.
https://doi.org/10.1007/978-3-030-64313-3_19
Précédent

- 207/443

Suivant