In contemporary linguistic theory, error no longer equates to equivocation (Corder, 1967; Corder, 1993; Selinker, 1972). Nowadays, errors are considered linguistic features that indicate the efficiency and effectiveness of the language teaching-learning process (Johansson & Hofland, 1994). The EA approach explores and explains why a given mistake is made at a precise moment in the teaching-learning process (Richards & Renandya, 2002). The possibilities show EA linked to the practice of foreign and second language teaching, turning this discipline into paramount for curriculum design success.

Thus, the ability to understand and deal with errors can benefit the teaching-learning process on several levels: assessing student performance, allowing educators to observe success and failure in the teaching-learning process, and exploring ways to implement more efficient curricula that may permit students to avoid making more errors and, on the contrary, produce better linguistic outcomes (Clerk & Rutherford, 2010; Macías, 2023).

One of the most notable contributions of EA is the setting up and theorization of the interlanguage hypothesis. Selinker presented it in 1969 when he spoke of "the provisional grammatical structure constructed by second language learners on their way to the target language" (as cited in McLaughlin, 1987, p. 69). However, William Nemser (1971), an expert in Comparative and EA, offered the most widely accepted definition of the concept of interlanguage conveying that "the learner's speech at a given time is the patterned product of a linguistic system distinct from the mother tongue and the target language and internally structured" (1971, p. 116).

Following Corder (1967) and Selinker (1972), McLaughlin (1987) describes the central cognitive processes of second and interlanguage learning on which, to some extent, the practical application of EA rests (Brown, 1994; Gass & Selinker, 2008): (a) language transfer and interference from the learners’ native language (L1); (b) overgeneralizations of rules in the target language, and (c) misconceptions of rules and usages in the target language.

CALL (Computer Assisted Language Learning) was introduced in linguistics and engineering studies in the 1950s; however, it took some time to become an effective teaching-learning tool (Chapelle, 2003). The basis of the CALL approach to second language teaching and learning is simple but compelling: Computers and computing devices can aid in second/third language learning if designed correctly (Levy, 1997).

The CALL approach remained a research field for some time before teachers, instructors, and learners used it in teaching and learning (Levy, 1997). Experts in the field claim that during the 1990s, the potential of CALL began to be applied (Chapelle, 2003). The underlying reason for the late use of CALL perspectives in teaching and learning can be found in the perceived lack of dissemination of the necessary technology to homes and educational institutions; schools, universities, students, and families had no access to computer resources in the early stages of the discipline’s development in the 1960s and 1970s (Schulze, 2011). It is often argued that the 1990s brought about the present technological revolution and facilitated the rise of personal computer-based learning systems (Chauhan et al., 2022).

CALL, as an approach and methodology, and AI applied to L2 studies have benefited from the theoretical background and academic debates, not only around computer science (Popovic & Ney, 2022) but also EA, error handling, and the interlanguage hypothesis (Jodaei, 2012). Recently, the effective integration of computer-assisted feedback in EA has been a fertile topic of discussion in the research field (Macías, 2023).

In this regard, recent research (Cabrera et al., 2014; Magid et al., 2022; Steadman & Kraut, 2018; Wobst & Lueg, 2022) claims that modern technology allows learners to practice and obtain feedback on their written and oral skills. It demonstrates that modern devices have incorporated effective technology required for testing processes. Therefore, CALL is a helpful tool for analyzing errors, as it contributes to identifying trends that ultimately lead to making mistakes (or avoiding them).

In the 1950s, one of the most notable personalities in applied AI, Alan Turing, defined what can be unequivocally understood as an intelligent system or device (French, 2000). Turing posed what is known today as Turing's imitation game: If a human listener cannot distinguish whether their speaker is a machine or a human, then we can speak of an artificially intelligent device/system/tool (Popenici & Kerr, 2017).

Concerning the contemporary situation for education studies, AI is a familiar concept and tool. One of the most striking recent contributions to integrating AI and education studies (linked to EA) is found in Luckin’s studies (2010). The expert claims that AI can support the development of technologies that enhance learning by minimizing process errors and maximizing the chances of positive outcomes. According to Luckin, AI-assisted models allow learners to work at their own rhythm and enable teachers and instructors to witness their students' processes from an external and objective perspective. It is also reported that the model helps teachers and instructors intervene whenever necessary and in varied modalities, not only face-to-face but also through distance tutoring (Underwood & Luckin, 2011).

Recent research (Kessler, 2021) affirms that modern technologies (especially those capable of processing and analyzing texts and oral productions) and experiences with the latest technological and communicative improvements (social media and virtual reality) allow teachers and instructors to effectively grasp the new dimensions of reality and the challenges of their students. Hence, Kessler (2021) argues that all the knowledge gained through these experiences helps teachers to produce a more individualized teaching-learning process by considering the specific cognitions and concrete needs of a particular learner. This has an overt impact on the feedback process: the more information the teacher can gather, the better knowledge they will have of the specific situation; it should result in better feedback that can have a dual, computer-assisted, and teacher-based nature (Chaudhry & Kazim, 2021). Thus, AI-based tools and data collection should help to develop better and more personalized teacher-student support and to achieve success in the learning process.

Today, it is undeniable that digital tools positively affect various educational processes and contribute to improving and boosting the system (Al-Fequi, 2012). AI applications, tools, and devices favor changing the roles played by educational actors and agents (education centers, teachers, managers, administrative staff, students, etc.). According to Dickson (2017), these new technologies will entirely and drastically change the interactional patterns between teachers/instructors and learners/students, as machines will offer an interactive educational solution to traditional problems. Dickson also states that these new technologies and AI tools can solve the problem of large-group interaction, which is especially important when providing feedback, improving student performance, and enhancing attitudes toward the teaching-learning process (2017).

Nowadays, teaching English as a second and foreign language benefits enormously from the introduction of AI, in particular, and Information and Communication Technologies (ICTs), among others. Language classrooms are very artificial spaces for learning, and communication and interaction often need to be improved. However, when these communicative and interactive situations are granted, learners can practice real-life skills (Haupin, 2016). As such, introducing technologies capable of simulating reality and forcing learners to communicate—imposing real-life simulated difficulties—can benefit the process as a whole (Dickson, 2017). These new apps, programs, and devices implement and enhance communication skills; through communication programs, conversational skills can be stimulated by introducing accurate, realistic, virtually interactive, and hands-on language training (Rabah, 2020).

Radwan (2020) has identified and described the various uses of AI to overcome difficult classroom situations and improve the teaching-learning process. According to him, the combination of educational tools and AI can:

• Be used to develop and enhance the ability to comprehend reading passages.

  Develop and enhance learners' translation skills using machine-assisted translation.

  Help learners improve pronunciation via automatic speech recognition tools.

  Help break down barriers for visually and hearing-impaired learners using text-to-speech tools.

  Improve writing proficiency through writing assessment techniques, registers, and automated handwriting correction tools.

On the one hand, the present study aims to evaluate and determine, preliminarily, the usefulness of introducing AI tools in teaching English to know the impact of these on written proficiency in a university course of English as a foreign language whose target level is B2 (upper intermediate). On the other hand, in a more practical approach to the discipline, we intend to explore the possibilities that AI offers for the feedback and assessment of learners of English as a foreign or second language. Our objective is to understand if AI helps the monitored learners to produce better writing (more appropriate to the pursued objective) and, therefore, if it improves the group’s teaching-learning competencies.

The identified problem to be researched constitutes the fundamental motivation of this study. Written competence and performance are often disconnected from daily practice in English as an L2, and this situation follows a recurrent and considerable delay in the feedback given to learners.

The starting point of this study implies that AI-based tools offer almost immediate and interactive feedback, and some of them enable and enhance the automation of feedback logs. Thus, we intend to demonstrate that combining EA with new intelligent technologies gives rise to possibilities for substantially improving the written production of language learners, particularly those learning English as a foreign and second language. It will be achieved by examining if there is a strong connection between these new tools and their feedback possibilities and improvements in students' written proficiency when assessed with AI-based tools.

This study attempts to answer the following questions addressing the described problems to fulfill the objectives:

• What are the most appropriate strategies and methods for employing AI in teaching/learning English in a university course?

  Is AI effective in developing different English teaching/learning processes and measuring their results?

  Is it feasible to use AI tools to improve written proficiency?

  Do learners perceive AI tools as a viable and enriching alternative to traditional feedback?

A group of 70 B.1.3 level English learners constitute the population of this research (performance and proficiency) in a single class at a public university in Madrid.

Students were asked to produce a text similar in number of words and topic to those they usually develop in the classroom; specifically, to write an opinion essay on a free issue of approximately 250 words, collected and anonymized by the teacher. Samples for this research were obtained by applying the spontaneous sample collection procedure, including collecting data from students when they focus more on content than grammatical style. Academic assignments and papers, tests, and essays can be taken as sources for data collection. Topics through the spontaneous procedure may include "personal information, future plans, religious, social and political issues, and the like" (Keshavarz, 2022, p. 80). The population participated 100%.

To meet the objectives, we designed a methodology based on a practical five-step method, which draws on the methodology of corpus and corpora linguistic studies. The following steps are essential to understand how this study was conducted and, if desired, to replicate it:

Step 1: creation of the first study corpus or Corpus 1. Seventy writing samples were collected from university English learners with an approximate proficiency in the target language B.1.3. The group members possess varying degrees of EFL proficiency, and approximately 65/70 have produced a language that meets the descriptors and criteria of the target level.

Step 2: division of Corpus 1 into study units. Once the samples are collected, they are first anonymized using labels to avoid possible interference (Kaiser et al., 2021; Zaki & Nosofsky, 2007), and second, they are randomly divided into two groups (35 writing samples in each group). These constitute the principal units of study.

Step 3: analysis of the study units. For each of these, one type of analysis is applied. One group (unit 1) is scanned via AI technology, capable of processing and evaluating texts at the grammatical and lexical level (Trinka.ai was used in this study). The second group was handled more traditionally: correction by the teacher with pen and paper but with the same level of analysis and detail.

Step 4: creation of the confirmation corpus or Corpus 2. This group is a mirror image of Corpus 1: seventy samples of written output by the same target population. Here, the main objective assesses (in a preliminary way) the possibilities offered by AI-based proofreading and text processing compared to a more traditional teacher-led proofreading system: It is essential to create a confirmation corpus that allows us to understand if and when errors are repeated. The analysis focuses on the repetition of and the nature of mistakes (Gass & Selinker, 2008; Rutherford, 2022).

Step 5: exploratory contrastive analysis of Corpus 1 and Corpus 2. Once Corpus 2 has been examined in detail, it is necessary to contrast it with the results of Corpus 1. This contrast allows us to inspect whether the AI minimizes the errors produced, as Kessler (2021) suggested.