Proof Assistants for Teaching: a Survey

Frédéric Tran Minh, Laure Gonnord, Julien Narboux · 2025

Read Proof Assistants for Teaching: a Survey by Frédéric Tran Minh, Laure Gonnord, Julien Narboux (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-13472.txt.

Introduction

Frédéric Tran Minh, Laure Gonnord, and Julien Narboux survey how proof assistants, software that checks mathematical proofs step by step, are being used to teach. Such tools have spread through mathematics and computer science research, and teachers have followed, using them in undergraduate courses on logic, mathematics, and computing. The survey gathers reports of classroom experiments with these systems.

It also covers software built or reshaped for the classroom: tutoring systems designed specifically for teaching proof, and general-purpose assistants fitted with friendlier interfaces or dedicated input and output languages. The authors organise this scattered experience so that newcomers can see which tools exist, how they have been adapted for students, and what instructors learned from using them. The result is a map of a young but growing teaching practice.

Catalog record

CREATORFrédéric Tran Minh, Laure Gonnord, Julien Narboux
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-13472

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Proof Assistants for Teaching: a Survey Frédéric Tran Minh, Laure Gonnord, Julien Narboux Abstract In parallel to the ever-growing usage of mechanized proofs in diverse areas of mathematics and computer science, proof assistants are used more and more for education. This paper surveys previous work related to the use of proof assistants for (mostly undergraduate) teaching. This includes works where the authors report on their experiments using proof assistants to teach logic, mathematics or computer science, as well as designs or adaptations of proof assistants for teaching. We provide an overview of both tutoring systems that have been designed for teaching proof and proving, or general-purpose proof assistants that have been adapted for education, adding user interfaces and/or dedicated input or output languages. Introduction The story of proof assistants originates in 1968 with Automath developed by De Bruijn [ Bru83 ] . Two categories of software can be identified. Automatic Theorem Provers (ATP) take a statement as input (often expressed in a language supporting First Order Logic) and act as a black box to determine if the statement is universally true or if a counter-example can be produced. CVC-5, Z3, Vampire, Superposition are some famous examples. On the other hand, Interactive Theorem Provers (ITP) check for validity formal proofs designed and written by a human user in a dedicated language, as expressive as possible (i.e., supporting higher-order logic (HOL) or dependent type theory (DDT)). In ITPs, some automation is provided to help the user fill in parts of the proof considered trivial by a human being. The proof state (current goal and hypothesis) is displayed at any point of the proof. Since the 1960s, numerous Interactive Theorem Provers, also called Proof Assistants, have been developed. Among the most famous still widely used nowadays are Mizar [ MR05 ] , Agda [ MN94 , Nor07 ] , Coq [ Coq85 , CB04 ] , HOL-Light [ Gor00 ] Isabelle [ NPW02 ] and Lean [ dKA+15 ] . All these systems mainly differ in the size of their kernels, the underlying foundations and the expressiveness of their language. 1 1 1 As the detailed description of each of these proof assistants is out of the scope of this paper, the reader can refer to [ Geu09 , NML+19 , Wie03 ] for surveys about proof assistants. These ITPs have been successfully used in mathematics to formally verify proofs containing computer code (famous examples are proofs of the four-color theorem by Appel and Haken in 1976, formally verified by Gonthier in 2008 [ Gon08 ] and the Kepler conjecture - Hales theorem, proved by Hales in 1998 following Toth in 1953 and formally verified by the FlySpeck project in 2006 in Isabelle [ Hal05 , HAB+15 ] ), and also in software verification (an example is the C compiler CompCert verified by Coq in 2016 [ LBK+16 ] ). Some work have been done to bridge the power and efficiency of ATPs with the expressiveness of ITP languages, by integrating calls to ATPs from within ITPs. Coq-hammer, SMT-Coq, Isabelle Sledgehammer, Lean-auto are some of them. At a time when Digital Media in Education are taking center stage in innovative pedagogy, this growing role for formalization in mathematics is meeting with the hope of the educational community that proof technology could offer renewed approaches to teaching proof [ BLT19 ] . At the crossroads between high school and higher education, the transition to formal mathematical language and formal proof is abrupt and many students struggle to adapt to the new expectations of their curricula : Hanna and Knipping review numerous research studies that have been devoted over the past 40 years to proof in education [ HK20 ] and summarize the expectations raised by proof assistants: (…) the use of proof technology may present a new opportunity for fostering mathematical understanding, similar to that which has already been seen in the use of other technology such as dynamic geometry programs. Interactive and automated theorem provers (ITP and ATP) are designed to provide a guarantee of correctness. The question is how and to what degree these tools can go beyond providing a guarantee and be useful in providing explanation as well. Focusing on students’ understanding of mathematics, and of proof in particular, while making use of automated proving, will be a promising avenue of research. Clearly, mathematics educators will have several challenging open questions to consider when looking ahead. However, in a further paper, Hanna and Yan regret the low consideration for these tools in undergraduate mathematics curricula and lack of research evidence that they actually improve learning and understanding [ HY21 ] . In this paper, we provide a survey of the different works related to the use of proof assistants for teaching. Even if automated theorem provers are important tools for interactive theorem proving, and can also be used in education for theorem discovering or checking statements (see the line of works related to GeoGebra [ BHJ+15 , HKR19 , KR20 , QS19 , Qua22 ] ), we will focus on the use interactive theorem provers for education. Section 1 exposes a variety of teaching experiences using proof assistants, from undergraduate mathematics to quasi-expert manipulation of program proofs. Section 2 describes adaptations that have been made to adapt the tools themselves, and also their user interfaces, to the special needs of education. We dedicate Section 3 to natural language interfaces. 1 Teaching experiments using proof assistants From the 70’s, with the first reported experiments using Mizar, to now, the literature is full of experiments that somehow reflect the existing status of proof assistants: their expressiveness, of course, and also the publicity that often comes from their usage in research developments. In this section, we report a variety of teaching experiments using teaching assistants, at different levels from undergraduate to master, and targeting various topics, from logic to more advanced formal verification of programs. 1.1 Teaching logic and meta-theory The first experiment of teaching (propositional) logic using proof assistants dates back to 1975, with the Mizar software. Matuszewski, Rudnicki and Trybulec developed an environment based on Mizar dedicated to teaching logic [ Try83 ] . This experiment, conducted at the Warsaw University, is described in the more recent paper [ MR05 ] , among other historical use of Mizar before 1989. A simplified environment (Mizar-MSE = MultiSorted with Equality), was used mostly to practice simple logical exercises, as the interface was judged austere. Similarly, other proof assistants have been widely used in the last 40 years as computer artifacts to support teaching logic-related topics such as propositional logic, induction, or set theory. The Coq Wiki [ The21 , The24 ] reports a list of such experiments with Coq. Teaching logic can rely on relatively low additional machinery as proof assistants expose natively their underlying logic. Exercising natural deduction, sequent-calculus is thus relatively straightforward, and numerous experiments have been conducted using general purpose proof assistant or ad hoc software. Using proof assistants to teach the meta-theory of logic is rarer. Some authors such as Villadsen and his co-authors have proposed using proof assistants with a deep embedding of some calculus, allowing formalization of meta-theory to teach both logic and automated deduction [ VFS18 ,

Read the complete text