expectations χ = ⟨χ1,…, χn⟩ (that is, distributions of
expected states in S after executing each sequential
action a1 ∊ π, starting in s).
The GDA model monitors π’s execution. This
involves four steps: ( 1) discrepancy detection, ( 2)
explanation generation, ( 3) goal formulation, and ( 4)
Step 1, discrepancy detection, compares the observations
obtained from executing action ai in belief state si with
expectation χi (that is, this tests whether any constraints are violated, corresponding to unexpected
observations). If a discrepancy d is found, then it is
given to step 2.
Step 2, explanation generation. Given next state s(i+ 1)
(provided by γ), si, and d, this process hypothesizes an
explanation e ∊ E (not shown in figure 2) of its cause.
Step 3, goal formulation. Given d, e, and s(i+ 1), this
process generates a goal gʹ ∊ G (not shown).
Step 4, goal management. Given a set (initially empty)
of pending goals GP ⊆ G and gʹ, this process may
update GP and will select the next goal g to feed to the
GDA does not specify what algorithms to use for
these processes or what representations to use for
these data models. Details of our initial GDA agent,
ARTUE, and its analysis are given in Klenk et al.
Several groups have used GDA as a starting point
for research on GR. For example, Molineaux and Aha
(2015) describe an abductive method for continuous
explanation generation that employs a constrained
heuristic search to identify plausible explanations,
given unexpected observations. 4 This method may
result in modifying initial state assumptions or the
action models of other agents in the environment.
Revised models can be learned and used to interpret
future similar occurrences (Molineaux and Aha
2014). Powell et al. (2011) use active learning techniques with GDA to acquire a function that maps
states to goals. Weber, Mateas, and Jhala’s (2012) EIS-Bot plays a complete real-time strategy (RTS) game
that uses GDA to select objectives (that is, which
units to produce). Jaidee, Muñoz-Avila, and Aha
(2013) describe a GDA variant that uses reinforcement learning to learn a goal selection function for
each unit type in an RTS game. Paisner et al. (2014)
describe how to model GDA in MIDCA. Dannenhauer, Muñoz-Avila, and Cox (2016) instead extend
Figure 2. A Depiction of the Goal-Driven Autonomy (GDA) Model of Goal Reasoning.
• Discrepancy Detector
• Explanation Generator
• Goal Formulator
• Goal Manager
UI d,e M π,χ s,g G