Justifying Machine Decisions through Argumentation and Semantics
Keywords: Explainable Artificial Intelligence, Knowledge Representation
and Reasoning, Computational Argumentation, Semantic Web
— Research Context —
Robots helping humans in performing their everyday activities are
becoming nowadays very popular, given the valuable impact they may bring
on society, e.g., robots assisting elderly people in their places to
support them in their everyday tasks. However, in order to concretely
interact with humans, intelligent systems are required to show some
human-like abilities such as the ability to explain their own decisions.
The research question we target for this post-doctoral position is “how
to explain and justify machine decisions to humans?”.
The domains are Artificial Intelligence, and more specifically Machine
Learning, Argumentation (as KRR formalism intended to explain decision
making) and Semantics (knowledge graphs).
This post-doctoral position grounds has for research context the ALOOF
project. In a nutshell, one of the goals of ALOOF is to equip autonomous
systems with the ability to learn the meaning of objects, i.e., their
perceptual and semantic properties and functionalities, from
externalized knowledge sources accessible through the Web. More details
may be found here: https://project.inria.fr/aloof/
— Business Context —
Accenture, as a service and consulting company, is bringing innovation
to its clients through various forms e.g., co-innovation through
workshops, rapid prototyping, piloting or R&D delivery. All journeys of
innovation starts by understanding our clients, their industry, value,
limitations and impact on the marketplace. Accenture The Dock in Dublin,
as the Accenture innovation hub in Europe, has been designed to showcase
the best of Accenture innovation and inspire our clients. The number of
clients attracted by The Dock and its visits has grown exponentially
over the past 12 months. Unfortunately not all client requests to
understand innovation in Accenture can be granted due to numerous
requests and preparation needed to showcase relevant technologies,
projects, innovations and people behind the innovation. Such a
limitation is due to the huge manual effort in preparing a client visit
and understanding its needs in term of innovation (in its particular
We aim at addressing this problem by developing machine-assisted
innovation touring. The machine, interfaced by Softbank Pepper, will be
responsible for guiding clients to relevant projects, team, people,
prototype, asset by understanding the client industry, its value,
limitations and impact on the marketplace. Data will be collected
internally to gather Accenture projects and assets, and externally to
consolidate it with the Web of data. The machine will be able to justify
its decisions (e.g., showcase of a project, team or asset) through
real-time interaction, ensuring a seamless machine to human (client)
journey across innovation in Accenture.
— Objectives —
The objectives of this postdoctoral position are:
– The gathering of bibliographical content to constitute a solid basis
for working on the following development.
– The definition of an Information Extraction module to extract
information from the raw data provided as input by the Accenture
company, and definition of a Semantic module able to construct the
knowledge graph based on the output of the Information Extraction module.
– The definition of a Decision Making module, so that given the goal of
the client and his own background, a plan is elicited to be executed,
i.e., the specific tour in the company building.
– The definition of an explanation module, such that the decision is
explained and justified to the clients by means of an argumentation
framework grounding on the generated knowledge graph.
– The scientific results obtained during the postdoc will be published
in top conferences and journals in Artificial Intelligence.
Accenture – INRIA
The candidate must hold a PhD thesis in Computer Science, with a
specialization on the Artificial Intelligence field. He/she must have
strong skills on the field and possibly in some frameworks and languages
related to it, knowledge on the Natural Language Processing field might
also help. An experience with Machine Learning frameworks is strongly
advised. Finally, he/she must have good English skills in writing and
Duration: 18 months.