Collaboration

AI is making students’ competency profiles compatible with job market needs

Mira Valkonen/ Kuva: Jonne Renvall
Describing competence is a diverse, complex and difficult task. Mira Valkonen is contemplating how to describe such skills as critical thinking and creativity. Photo: Jonne Renvall
In the Tampere Universities community, an artificial intelligence-based tool is being tested with the aim of identifying individuals’ competence and comparing it with the competency needs of working life.

“In theory, we could use artificial intelligence (AI) to discover whether we are providing students with the necessary skills they will need in the labour market,” says Specialist Mira Valkonen from Tampere University’s Lifelong Learning Services.

Valkonen is involved in a five-year continuous learning project that is piloting a tool made by a company called Headai with students at Tampere University and the Tampere University of Applied Sciences (TAMK).

Based on various textual data sources, the tool compiles the students’ competence profile, which is a set of words describing their competence. Students can compare their competence profile with job advertisements and assess the suitability of their competence for the regional job market as well.

The tool will also include a section serving the needs of companies and individuals, in which proposals for competence development will be compiled. The development work is carried out in co-operation with the ESF-funded Osaamista menestykseen project of Education Manager Teemu Rauhala and Senior Research Fellow Jussi Okkonen.

The project involves small groups of students from different fields: ICT students, social research students and business studies students. The ICT students come from Tampere University and TAMK. The social research students come entirely from the University and the business studies students from TAMK.

The aim of the project is to reform higher education so that it will better meet the changing needs of working life and society. The project will be funded by the Ministry of Education and Culture in 2019–2020 and continue until the end of 2023.

Varied lists of terms of expertise

In the first meeting with the students, the AI-based programme picked competence-related words from degree data that was based on the students’ curriculum. After this, the students entered their CV in the tool. The aim was to test how well the machine recognised skills in the various descriptions of the degree programmes and CVs.

The preliminary observations from the first meetings showed that AI produced a great deal of words describing competence from the degree data of the different students. Their aptness and relevance varied.

According to Valkonen, the competence word results of ICT students were reasonably relevant because the curricula of the ICT field have been written in a form that enables artificial intelligence to recognise the words that describe competence.

“We got clear and good lists of words. The tool also suggested other words that might be connected to the topic,” Valkonen notes.

In social research, the lists of competence words were considerably more varied than the other lists, but the words that were found did not make much sense. However, Valkonen found the lists interesting and long.

The word lists describing the competence of business students were the most interesting. The words were not self-evident, but most of them made sense.

It is hard to describe competence

Valkonen started to consider the added value of words in identifying competence.

“We need to look critically at what we want to use the word lists for and what individual words tell us in the end tell. We can list words in a CV without artificial intelligence, and the words themselves do not necessarily bring much added value, at least from the perspective of supervising studies. The most important thing is the individual’s reflection and own understanding of the skills that is created in the process and in which the tool can help,” Valkonen says.

On the other hand, a single competence word can serve as a good search function when searching for specific content in a large body of text. So far, the usefulness of individual words for identifying competence is limited. Therefore, the model based on word lists still requires much development. The challenge already arises in deciding what is a good word to describe different skills.

“Describing competence is a diverse, complex and difficult task. Describing generic skills is particularly challenging. How can we describe such skills as critical thinking or creativity, let alone verify the extent to which an individual has such skills?” Valkonen asks.

Programmes using artificial intelligence are constantly learning new things as they build word lists from openly available materials. Headai’s algorithm system currently has approximately 40,000 relevant competence words and an estimated 90,000 relevant words. Valkonen points out that a pilot version is being used at the start of the project, which is constantly being developed.

Conflicting job ads and labour market needs

The basic idea of the project is to find new ways to identify the skills of individuals and the skills needed in society, so that, for example, students and employers can find each other more easily. It will also help students to develop their skills in the right way. In the future, technology could serve not only the individual but also, for example, the team level of organisations.

In Headai’s application, job market needs are described by job advertisements that come from the public te-palvelut.fi and monster.fi websites. The tool first extracts the competence words from the students’ CV and curriculum and looks for equivalents in the job advertisements.

According to Valkonen, describing the job market needs based on job advertisements is a critical and even ethically problematic phase.

“Job ads do not fully reflect the skills needs of working life. Recruitment is increasingly taking place through direct searches and networks, in which case vacancy ads may not be published at all. What if the tool does not suggest any suitable jobs for a student? Such situations give rise to many ethical questions,” Valkonen points out.

Estimates about the future suggest that increasingly many complex skills – such as creativity, problem-solving and empathy – are needed in working life. The ability to identify these skills or to describe them in job advertisements is usually lacking in organisations even when such skills are required.

Valkonen believes that recruitment practices will develop so that, in the future, the number of traditional job advertisements will decrease even further.

“Therefore, it is necessary to examine whether the data we select now is a reliable source and whether they are saying anything about what is needed in jobs or saying something entirely different. However, it is hard to find other data. That is why it is important that we start experimenting and developing these possibilities for identifying and finding skills from our existing data sources,” Valkonen says.

“It would be great if we could combine different sources of information as well as weak signals about the skills that will be needed in the future. The coronavirus is a good example because it is an unforeseen thing that has generated demand for new professionals nearly overnight,” Valkonen mentions.

Artificial intelligence to help individuals and teams

Competency profiles could help build functional teams at workplaces. They would enable employers to find new employees among university students more easily, and students could find mentors, for example. The project is developing such a digital marketplace for competencies with the help of the Bazaar concept, led by Specialist Katariina Yrjönkoski from Tampere University’s Lifelong Learning Services.

“The possibilities are endless. The long-term vision is that we will find other ways of identifying competence than degrees and credits,” says Valkonen.

All visions are not realised instantaneously. Valkonen sums up the experiences gained so far by the fact that identifying competence has turned out to be much more complicated than originally thought. According to her, it is important to first determine what the information obtained about competence is to be used for.

“We cannot say measurably and unambiguously what kind of competence or how much of it a person has. There is no completely comprehensive way to identify competence. If such a method had been invented, the inventor would already be a millionaire,” Valkonen points out.

However, even if there is no objective way of measuring competence, subjective assessments can also be useful.

“We can use AI-assisted competence data to help individuals and support competence development in very diverse ways. There may be self-assessments and co-worker assessments. In such cases, it is irrelevant whether the assessment is objective or subjective because the comprehensive picture will help to understand one’s own skills,” Valkonen says.

Valkonen considers it a more difficult task to find a uniform way for universities to describe competence, but different solutions are being sought for that as well.