Logic-Based Methods for Verifiable and Explainable Artificial Intelligence

Dear Colleagues,

Notwithstanding the recent development in Artificial Intelligence, current algorithms are relatively untrustworthy. When making decisions based on big data,  the white-box reasoning behind it as well as the possible consequences that a decision might entail are often opaque, and this has ethical concerns in particular. This is why decision process systems using classical AI might become unjust, harmful, and deeply unequal, thus badly affecting  the population at large. On real-world big-data, where data veracity is also questioned, is often debatable whether automated extraction processes or people's opinions  are trustworthy, thus requiring solid verification and a knowledge base of trustworthy data for external validation. The early adoption of logical-driven machine learning models might help in these respects, allowing to directly extract a formal specification providing a machine-readable explanation of either the underlying learning process or  a human-readable model that can be further validated by a domain expert of interest.

Current approaches providing machine learning explainability (Shapely Values, LIME) often  rely on a post-hoc explanation, for which the input data is correlated to the  outcome. These have the shortcoming of  providing no easily readable explanation of why the results should hold, nor ensure the correctness of the underlying machine learning problem, as they usually provide an approximation of the learning model. This postulates that explanations  within the learning process might be a helpful complement to  post-hoc approaches. In particular,  explainable learning approaches that are logically driven, such as abduction reasoning, or that can provide an explicit logical model explaining the classification outcome, such as specification mining and learning techniques are desirable. Moving from classical AI to logical-driven, it is also essential to provide certification guarantees on the correctness of the learning process while also studying the formal properties of such frameworks, thus providing a better understanding of the actual capabilities of current black-box learning algorithms. Another important element towards logical-AI is the early adoption of Ethical-driven approaches in the form of both human-readable and machine-readable common-sense reasoning.

We invite contribution to this Special Issue on Verified Artificial Intelligenc,  that we hope will better shape the future of both Logic and Artificial Intelligence. The Special Issue aims to broaden the boundaries of logic in AI by also considering related topics  (e.g., ethics), as well as the application of logic in other domains (temporal reasoning, fuzzy logic, quantum logic).

In this Special Issue, original research articles and reviews are welcome. Research areas may include (but not limited to) the following:

• Relational Learning and Deductive/Abductive Reasoning;
• Theoretical Computer Science results and properties for Artificial Intelligence Algorithms (Reinforcement Learning, Deep Neural Networks);
• Learning with Logical and Correctness Guarantees (Fuzzy Logic, Quantum Logic, Modal Logic, MV logics and more);
• Verified Artificial Intelligence (Specification Mining, Specification Learning, Hybrid Explainability);
• Explainable AI;
• AI with Ethical Guarantees and its Applications (e.g. Healthcare, Law and Jurisprudence).

We request that, prior to submitting a manuscript, interested authors initially submit a proposed title and an abstract of 200-300 words summarizing their intended contribution. Please send it to the Guest Editor, or to the Assistant Editor of Logics. Abstracts will be reviewed by the Guest Editors for the purposes of ensuring proper fit within the scope of the special issue. Full manuscripts will undergo double-blind peer review.

We look forward to receiving your contributions.

Dr. Giacomo Bergami
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Logics is an international peer-reviewed open access quarterly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 1000 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• logics for AI
• theoretical computer science results from artificial intelligence
• relational learning
• logic and learning