Early Stage Research Position available for the H2020 MSCA- NL4XAI project

One ESR PhD position in Artificial Intelligence within the framework of NL4XAI: Interactive Natural Language Technology for Explainable Artificial Intelligence, a project funded by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie Grant Agreement No. 860621.

NL4XAI is a European Training Network (ETN) project, which will train 11 creative, entrepreneurial and innovative early-stage researchers (ESRs), who will face the challenge of making Artificial Intelligence (AI) self-explanatory and thus contribute to translating knowledge into products and services for economic and social benefit, with the support of Explainable AI (XAI) systems.

The focus of NL4XAI is in the automatic generation of interactive explanations in natural language, just as humans naturally do, and as a complement to visualization tools. As a result, ESRs are expected to leverage the usage of AI models and techniques even by non-expert users.

Reference number: NL4XAI- ESR4

PhD research topic: Explaining Logical Formulas

Objectives: Logical formalisms play an important role in many areas of study, including Artificial Intelligence, Philosophy, and Linguistics. Yet students and users of logic sometimes struggle to grasp the exact meaning of logical formulas. This happens when formulas are very complex, or when they have an unusual structure (for example when they are produced by a computational theorem prover). It also happens when learners are not yet fully accustomed to the conventions employed by the logic (e.g., the meaning of connectives in the so-called paradoxes of material implication).

To investigate how Natural Language Generation (NLG) techniques can be employed to automatically and effectively explain logical formulas to non-experts using texts that are formulated in ordinary language (e.g., in English, Dutch, or Chinese), the candidate will investigate computational techniques for simplifying and translating logical formulas into optimally intelligible NL text, and empirically evaluating the usefulness of the resulting text for users.

A key task is to convert a given formula into a form that makes it intelligible to people, for example by removing parts of the formula that follow from background knowledge, and by exploiting particular strengths of Natural Language. But Natural Language pose problems as well: an important challenge is to minimize textual ambiguities (for example, when a sentence is produced that can be parsed in two different ways). The choice of logical formalism is still open; promising options include First Order Predicate Logic and formalisms in Formal Argumentation. The choice of natural language is open; promising options include English, Dutch, and Chinese, which are all studied in the department.

Project results may include: (1) Algorithms for finding the simplest formula equivalent to a given input logical formula. (2) Algorithms for expressing the content of a (simplified) logical formula in NL text. (3) An interface that allows a user to enter an input formula, and to choose between different simplification/presentation strategies. (4) Evaluation results regarding the effectiveness for human users of the above techniques. One option for evaluation is to test whether the generated translations are understood correctly by users; another is to find out to what extent the generated translations help the user’s overall learning of the logical formalism.

The proposed research combines aspects of computational logic, language, and experimental psychology. Since only few people master all three of these research areas, the precise focus of the project will be chosen in conversation with the successful candidate. The supervisory team will include Kees van Deemter (of Utrecht University) and Jordi Levy (of CSIC’s Research Institute for Artificial Intelligence in Barcelona).