Natural Language Explanations in Human-Collaborative Systems

Scalise, Rosario; Rosenthal, Stephanie; Srinivasa, Siddhartha S.

doi:10.1145/3029798.3034809

Cited by 5 publications

(5 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…1 Similarly, participants seemed to reliably agree with the claim 'If I do not know that an AI's internal algorithm is working properly, but that AI is able to explain why it makes decisions, I trust that AI' (average response on the Likert scale was 3.25). 2 Thus, as far as the (admittedly limited and preliminary) evidence collected through our survey suggests, for the purpose of generating and sustaining trust in AI-mediated recommendations, these systems would benefit from the ability to explain why they made their decisions, similarly to how a human being would (Kunkel et al, 2019;Scalise et al, 2017). So, what kind of explanations do humans expect from AI?…”

Section: Functional Explanations and Reasonsmentioning

confidence: 97%

What kind of Explainable AI do users need?

Buccella,

Chen,

Linstead

et al. 2023

Preprint

View full text Add to dashboard Cite

Artificial intelligence (AI) systems that make predictions, translate text, generate text and images, and much more are ubiquitous in modern life. Such AI systems typically perform tasks that humans cannot perform with the same degree of speed and efficiency, so they are increasingly integrated into human decision-making. But they also tend to be black boxes, in the sense that even experts in machine learning do not understand how they arrive at their output from their given inputs. In this paper, we argue that to promote positive AI system-user interactions, such AI systems should be equipped with a user interface that provides reasons for the systems’ decisions. By ‘reasons,’ we mean linguistically articulated explanations of why a certain output was issued, in a similar sense to what we would expect humans to provide as explanations for their actions. Some preliminary survey data suggest that non-expert users generally highly value being able to have one’s ‘why-questions’ answered directly by the AI. We further propose that this feature will increase non-expert users’ trust in AI moving forward, which is desirable as AI takes up an increasingly larger role in our society.

show abstract

Section: Functional Explanations and Reasonsmentioning

confidence: 97%

What kind of Explainable AI do users need?

Buccella,

Chen,

Linstead

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Explanations are a core component of human interaction (Scalise et al, 2017;Krening et al, 2017;Madumal et al, 2019). In the context of Machine Learning (ML), explanations should articulate the decision-making process of an ML model explicitly, in a language familiar to people as communicators (De Graaf and Malle, 2017;Miller, 2018).…”

Section: Related Workmentioning

confidence: 99%

A Study of Automatic Metrics for the Evaluation of Natural Language Explanations

Clinciu¹,

Eshghi²,

Hastie³

2021

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume

View full text Add to dashboard Cite

As transparency becomes key for robotics and AI, it will be necessary to evaluate the methods through which transparency is provided, including automatically generated natural language (NL) explanations. Here, we explore parallels between the generation of such explanations and the much-studied field of evaluation of Natural Language Generation (NLG). Specifically, we investigate which of the NLG evaluation measures map well to explanations. We present the ExBAN corpus: a crowd-sourced corpus of NL explanations for Bayesian Networks. We run correlations comparing human subjective ratings with NLG automatic measures. We find that embedding-based automatic NLG evaluation methods, such as BERTScore and BLEURT, have a higher correlation with human ratings, compared to word-overlap metrics, such as BLEU and ROUGE. This work has implications for Explainable AI and transparent robotic and autonomous systems.

show abstract

“…Detailed technical logs of a robot's experience were originally available only to trained researchers (Landsiedel et al, 2017;Scalise et al, 2017). Recent work, however, has generated natural language descriptions of a robot's travelled path from them.…”

Section: Related Workmentioning

confidence: 99%

Plan Explanations that Exploit a Cognitive Spatial Model

Korpan¹,

Epstein²

2021

Proceedings of Second International Combined Workshop on Spatial Language Understanding and Grounded Communication for Robotics

View full text Add to dashboard Cite

Ideally, people who navigate together in a complex indoor space share a mental model that facilitates explanation. This paper reports on a robot control system whose cognitive world model is based on spatial affordances that generalize over its perceptual data. Given a target, the control system formulates multiple plans, each with a model-relevant metric, and selects among them. As a result, it can provide readily understandable natural language about the robot's intentions and confidence, and generate diverse, contrastive explanations that reference the acquired spatial model. Empirical results in large, complex environments demonstrate the robot's ability to provide humanfriendly explanations in natural language.

show abstract

Natural Language Explanations in Human-Collaborative Systems

Cited by 5 publications

References 9 publications

What kind of Explainable AI do users need?

What kind of Explainable AI do users need?

A Study of Automatic Metrics for the Evaluation of Natural Language Explanations

Plan Explanations that Exploit a Cognitive Spatial Model

Contact Info

Product

Resources

About