Artificial Intelligence as a Mentor for Lifelong Learning

Background and Problem

The transformation of society into a knowledge society and ever faster technological progress create a strong need to keep vocational skills and knowledge up to date and to extend them, in the spirit of lifelong learning, even after initial vocational training. Owing to the rapid pace of technological progress, the ever-broadening knowledge base and the wide variety of continuing training provision, it is becoming increasingly complex for those interested in continuing training to identify and choose suitable courses for themselves. Furthermore, it is difficult for the individuals concerned to foresee how future-proof particular training content and the competences they are aiming for will prove to be in a labour market that is constantly evolving.

The collaborative project ‘KIM’ therefore aims to put those interested in continuing training centre stage and to offer them targeted support in choosing their continuing training provision. To this end, users select a career goal and upload their CV to the project website; on the basis of the CV, the AI determines which skills have already been acquired and which still need to be developed. Users can then confirm these or amend them manually. Using this information, precisely matched continuing education and training courses are identified and proposed for acquiring the skills that are conducive to or necessary for the desired career goal. Trend analyses of the labour market and of research also feed into the search for suitable courses.

Research question

• Which criteria contribute to users’ acceptance of AI?

• Which criteria influence the subjectively perceived fit of the continuing education and training courses proposed by the AI?

• How can sustained use of the application be achieved?

Approach

For the overall solution to be developed, it was essential to use occupations, the associated competences and qualifications as consistently as possible within a single, uniform taxonomy. Following a corresponding analysis, it was decided to draw on the ESCO classification for this purpose, as it offered a comparatively large database of occupations and the competences assigned to them. ESCO also provided a data interface through which the required content could be integrated into the project’s own solution in its most up-to-date version at any given time. However, instead of using the more than 14,000 competences listed in ESCO, a higher level of abstraction was adopted, which made it possible to reduce these to around 2,000 competence areas. This step facilitated the manual training of the AI for assigning CV data to the competence areas and reduced the effort involved in manually matching course content. The CVs required for training the AI contained sensitive data. To ensure that the work complied with data protection requirements, particular attention was paid to data protection throughout the entire project and a corresponding data protection concept was established. The CV data were broken down into individual sentences so that no complete CVs were stored. Instead, they were stored locally on the users’ computers and only uploaded to the database after the personal data had been filtered out manually. These measures helped to guarantee data protection standards in the project while at the same time allowing the data required for training the AI to be collected and processed in a secure and data-protection-compliant manner. The data sets needed to train the AI consisted of CVs that had been made available as well as public CV data from the Federal Employment Agency (Bundesagentur für Arbeit), which covered a broad range of occupational categories. To capture user acceptance, the individually perceived fit and the long-term usage behaviour of users, qualitative and quantitative studies were conducted with potential users and evaluated. The results of the surveys were able to feed into further implementation. The output of the project was an application that presented users with continuing education and training curricula precisely tailored to their career aspirations, in the form of various courses that supported the acquisition of one or more of the skills they were aiming for. Users could select and book these courses directly.

Results

The results of the survey on user acceptance showed that a transparent data protection concept contributed to trust in an AI-based application just as much as the opportunity to turn to real people as contact persons. Being able to watch a video in which real people described the key facts about the application also contributed to user acceptance. The results of the study on the perceived fit of the continuing training curricula proposed by the AI application further showed that AI acceptance and the perceived added value of using the application had a positive effect on the perceived fit, as did user-friendliness, the trustworthy impression made by the application and the outputs of the back end. These results were consistent with the findings of the final survey on willingness to use the finished application on a lasting basis. Users wanted the application to be presented in an intuitive and purposeful way. Participants described using the application as having ‘great entertainment value’, with the AI itself being perceived as fascinating. It was said to deliver ‘out-of-the-box’ ideas that were inspiring and sparked interest. The proposed courses were reported to be more extensive and more varied than an independent search would have yielded. With regard to the number of continuing training suggestions, a condensed ranking, optionally accompanied by a percentage indication of fit, was proposed as an improvement. In the interests of greater efficiency, it was also suggested that the titles of the individual courses be made more precise.

Publication

Gmeinder, Svenja; Petersen, Maren; Kuehne, Maxim (2026): Transfer digitaler Innovationen in der beruflichen Weiterbildung: Akzeptanzfaktoren einer KI-gestützten Mentoring-Anwendung zur Förderung lebenslangen Lernens. In bwp@ 49: Innovation und Transfer in der beruflichen Bildung, 1 - 22