Ontological foundations for contrastive explanatory narration of robot plans

Abstract

Mutual understanding of artificial agents' decisions is key to ensuring a trustworthy and successful human-robot interaction. Hence, robots are expected to make reasonable decisions and communicate them to humans when needed. In this article, the focus is on an approach to modeling and reasoning about the comparison of two competing plans, so that robots can later explain the divergent result. First, a novel ontological model is proposed to formalize and reason about the differences between competing plans, enabling the classification of the most appropriate one (e.g., the shortest, the safest, the closest to human preferences, etc.). This work also investigates the limitations of a baseline algorithm for ontology-based explanatory narration. To address these limitations, a novel algorithm is presented, leveraging divergent knowledge between plans and facilitating the construction of contrastive narratives. Through empirical evaluation, it is observed that the explanations excel beyond the baseline method.

Figures (3)

• Figure 1: Proposed approach overview and its application to a prototypical human-robot interactive scenario that requires contrastive explanations. A human gives an ambiguous command, a robot generates different plans, stores knowledge about them, and contrasts them and reasons about which plan is better to execute. Finally, the inferred knowledge is used to explain the rationale.
• Figure 2: Graph visualization of the proposed ontological model. The defined plan's properties are subclasses of Quality, indicated by the only continuous arrows. The graph depicts the relations between plans and their properties (e.g., 'has cost', 'is cost of'). It also shows multiple relations holding between plans (e.g. 'is cheaper plan than'), and three relations between qualities (e.g. 'has better quality value than').
• Figure 3: Graphical representation of the different levels of specificity (S1, S2, S3) and their respective depth in the knowledge graph. Nodes in black correspond to initially retrieved tuples that are later pruned because they are non-divergent. S1 contains direct relations between the ontology instances to be compared (e.g. Plan A is cheaper than Plan B). S2 compares how the target instances relate to other objects, and the relations between those objects (e.g. Plan A has a specific cost (Cost A) while Plan B has a different cost (Cost B), or Cost A has better quality value than Cost B). S3 goes deeper in the knowledge graph, comparing the relationships of the objects of the previous level with other ontological objects (e.g. Cost A has data value 27).

Theorems & Definitions (3)

• Definition 3.1
• Definition 3.2
• Definition 3.3