The changing nature of work
A lot of discussion is happening about how work is changing. Digital tools are becoming increasingly present. Employees need digital skills in order to get their work done. The definition of digital skills is evolving together with the opportunities and challenges digital skills provide. It began with digital skills referring to ICT skills and literacy, being able to operate technological tools specifically computers. Here the focus is on using computers to get work done. However, this has evolved to a broader concept, namely the “confident and critical use of ICT for work, leisure, learning and communication”. At least this is now the definition used by certain institutes such as the European commission. This definition is broader, as next to the ability to use ICT technologies, it includes cognitive skills and attitudes towards technologies. Kirsti Ala-Mutka describes digital skills as the “motivation and competence to strategically and innovative apply tools in different context” (p. 4). In short, digital skills are not anymore what they were when I was in high school, knowing how to turn a computer on, open word, and write a text. It goes beyond the operation, and implies knowing when to use what tool to get the necessary information in the most appropriate format to complete a task.
Are digital skills part of employees human capital?
For companies digital skills are important as they are seen as a competitive advantage. Research done by Capgemini shows that most companies see the importance in digital skills, but lack employees with digital skills.
To get value out of (technological) tools, it is necessary that people can use them. Hence the necessity for developing digital skills in employees. Unfortunately, as the research by Capgemini shows, that seems to be lacking in many employees. If a skill is lacking, training to develop the skill is the answer. There was a time when, according to economic labor market theory, companies only invested in company-specific skills. The rational reasoning was: Why invest in training for our employees if they can take the knowledge somewhere else. If this is still relevant (and the emphasis is on if), employers will not invest in digital training skill. But this is not my argument and my line of research.
I’m not investigating why only 46 % of employers invest in developing digital skills. As a researcher at a business school I consider it part of my job to equip graduates for their work. Of course this does not mean that graduates are perfectly prepared for their first position. But we, the body of researcher and instructors at business schools should at least give them enough knowledge and skills to be able to learn and grow in their role.
Now, for the sake of argument, let’s assume that it is the business schools’ job to equip students with digital skills. This means that we, the instructors, need to master the skills. Theoretical knowledge is good, but not sufficient. That means, that we need to have digital skills.
Digital skills in Academia
What is the situation with regard to academics level of digital skills? Well, before researchers conduct studies we read what others have done in the field. The idea is to build on the work of others. That would mean that academics are good at looking for information and defining problems, which is included in the definition of digital skills.
But what about using tools, and not just Adobe, SPSS, STATA, NViVo, Office products, and Endnote. These are core tools to read information, analyze data quantitative and qualitative data, write articles, and organize references.
What about newer tools? Those that come from the open source community or those that help dealing with a specific problem, such collaborating with researchers who work at different universities or analyzing large data sets. More concretely, I was interested what place tools and skills necessary to analyze large data sets have taken in academia. With that question in mind I was looking for data that could answer it.
Matthew Salganik teaches in his book Bit by Bit that before looking for data researchers should describe their ideal data set. The ideal data set would contain information about all the skills from all the researchers. Now this is possible, but not realistic. It requires a lot of scrapping of websites, collecting CVs, and scanning them for information about skills. An alternative is to analyze vacancies. The assumption is that by analyzing the description and requirements of academic vacancies I am able to get insights into what skills are searched for in new hires. From a strategic perspective, academic institutes could look for two things in new hires based on their strategic agenda: First they could want to strengthen their current expertise position (‘market position’) by looking for hires who have an overlap in skills. Second, they could want to diversity their position by building up new research streams. In that case they would look for candidates who have complimentary and different skills. I was fortunate to be able to collaborate with Academic Transfer, a company publicizing academic vacancies in the Netherlands.
Analyzing academic vacancies
To explore the data set of vacancies, we decided to apply natural language processing, specifically Latent Dirichlet Allocation (LDA), to discover the topics across the many vacancies. My student assistant worked hard to preprocess the data (cleaning, subsetting, stemming, and filtering out stop words) and run various topic models to find the ideal number of topics to describe the collection of texts. We decided to add context-specific stop words to exclude words that were too general. As I’m doing this work for a conference submission that is due next week, this step could be refined, but right now it’s good enough.
Very briefly, natural language processing is the task of asking a computer to analyze text, known as unstructured data, to find patterns in it. LDA is one method that can be applied to detect models. LDA looks at all the words in a text across all documents and then assigns a probability that word A is in topic 1, word B in topic 1 and so on until for all words a probability is calculated that they are part of a specific topic. This is done several times.
From all the topic models my student assistant run, I selected the one that made the most sense. Yes, this step is subjective. I looked at the words for each topics and what this topic could be about. From that the student created a matrix containing the vacancies in rows and topics in columns. A cell contained the probability that a vacancy is about a specific topic. Here a short example:
The vacancy with this processes (stemmed) requirement text “candid are expect to have an msc in econom traffic engin or a compar field and to have a interest in analyt and numer model profici in english in speak and write is essenti the postdoc has a doctor in one of the abov field” belongs with 81 % chance to topic 2 (computing skills) and 7 % to topic 5 (evidence of research skills). The other percentages are spread across the other four topics.
While the fun part still needs to be done, analyzing the matrix, looking at all of the generated topic models already provided some interesting insights:
1. There is a lot of turnover or recruitment in life science research. A lot of vacancies had words referring to microbiology, and gene research. This could also indicate an uptake in life science research in the Netherlands.
2. You can recognize a vacancy from Maastricht University by its requirement section. Depending on the number of topics that we asked the computer to use, it gets its own topic, by grouping problem-based learning and small tutorials into its own topic.
3. Institutions use vacancies try to position themselves in a specific way (‘ambitious career’).
4. Teaching in the social science seems to have different requirements than teaching in the life science.
5. Working in academy means research and teaching across all disciplines.
Over the next days I’ll be doing social network analysis on the vacancy-matrix to find out more about digital skills requirement for working in academia. Look out for the next post by subscribing to my blog.
If you like to know more, or work with me, please contact me.