18 months ago, I found myself without a job, not only that, I also had no replacement income. So getting a new job was really urgent. At that point I saw myself as a global award-winning learning expert, an established PhD person, as well as an innovator, I thought that with my qualifications a headhunter would certainly contact me and offer me jobs “We need you Inge!” … well, that didn’t happen [with gestures and in fun] [mad typing on computer] So I started to look for a job on the internet: I updated my CV and started screening jobs on LinkedIn, on Academia, on SteppingStone, … each day I would spend hours looking for jobs. And if the jobs didn’t match, I would spend hours figuring out how I could reskill myself by learning for those jobs. But my skills simply didn’t add up, and the learning needed was unclear when reading the job information. It became really frustrating [pause after gesture]
So, I started dreaming. What would make it easier for me to find my ideal job? It should be a job with most of the matching skills, so I only need some additional training … what would make this search easier? [Pause, mimic thinking posture] Just imagine that you only had to open a computer, you type in your ideal job, you press enter … and a list of jobs are provided that really match your skills and they would even point to courses that would enhance the skills you need! Waw. I started to draw up options, challenges, roadblocks… just for fun… Hey, I was without a job.
Right about that time I met up with an old colleague of mine. He asked me whether I had been able to find a job. I said no, but that I did have an idea to find jobs more easily. As he listened to my idea, he told me that he had just launched a similar project, but that he was missing a learning expert. So, he immediately hired me to become part of the existing team… I was headhunted after all! [enthusiastic posture and gesture]
Now let me tell you about the project my colleagues and I are working on. The skills engine. We all know search engines, but these search engines can be used to find anything you like. In this case we want to build a skills engine, an engine that can provide you with real job options, as well as courses that will enhance your skills and make it easier to land this job.
To get the skills engine up and running a specific Artificial Intelligence process is used. It is called Natural Language Processing.
But what is it? Basically, Natural Language Processing is Human linguistics on speed. We - as humans – use information which is transmitted via language to construct meaning. Let me give you an example [slide with words love, lips, hugs, cosy, you
If a person says: love, lips, hugs, cozy, you… What do you think he or she means? [audience]
Indeed, we deduct from there that the other person wants something intimate from us. Humans don’t need full sentences to get a hint of what someone means. We have acquired the capacity to understand relationships between words. Words are clustered in our minds to concepts, and these concepts are interrelated. So where does natural language processing differ from what we humans do with language? Natural language processing works on a much bigger scale – a big data scale – and it analysis language information at a much higher speed.
Such a Natural Language Processing -engine was built and set up.
Now we needed the right field to test our engine. We needed a niche area, and luckily InnoEnergy is niche enough and thriving with innovation, so it is ideal to test out the skills engine. InnoEnergy – the company I work for – provides courses and trainings for the renewable and sustainable energy sector. Well, this is niche enough [supporting mimic], on top of that the green energy sector is crucial for our sustainable earth, we need to be able to reskill employees at a much bigger rate.
The AI approach was chosen, InnoEnergy is a perfect field to test a pilot. Now, we just need to come up with a reskilling and upskilling plan that would make our skills engine useful.
The first step was to grasp where an industry was going. But how do you know where your industry is going? [slight pause] In the past we turned to industry goeroes (like Jack Ma) to give us future insights, but these predictions were hardly based on facts. Now with Natural Language Processing we can analyse vast amounts of texts in just a bit of time, based on the facts from these reports. So we feed the machine masses amounts of industry reports. We provided it with white papers, government reports, industry reports…. Then we compared one year to the next. If the analysis of 2016 reports showed the major focus was on solar energy, and the analysis of 2019 papers indicated that the major focus was on wind energy… then we knew that the upcoming skills were more related to wind energy than to solar. Now we had a tool into place to see where an industry was going.
Next step: the skills. If we know where an industry is going, we need to know which skills are linked to that transition. So we looked at different skills taxonomies. For each job, you need to know the necessary skills. It is not enough to know which skills are needed, you need to know which learning outcomes precede these skills in order to provide training or learning options to get to these skills. After an intense search we came up with the right skills taxonomy, including a description of the knowledge that was needed for each skill.
Step 3, which skills do people have? If you are 40 or 50 years old, a 4-page resume scarcely reflects all of your actual skills and capacities. Though we keep sending them to HR people who don’t know us either, but have to evaluate us on the basis of these short CVs. Luckily, now with AI we can analyse cv’s, resumes, … and understand the directly mentioned skills, but also the non-listed, assumable skills present in this person thanks to Natural Language Processing. For Natural Language Processing can find underlying relations from textual concepts.
In Step 4 we look at skills gaps and possible courses. What if there is an ideal job, but you are missing some key skills? Thanks to step one in which we analyse industry reports, we know which skills are emerging. Thanks to step 3 we know which skills a person has, so now we only need to compare both to see which skills a person is missing for these future jobs. The skills gap. Now we need to find the right courses that will address this skills gap … We do this by crawling the internet and looking at course descriptions, from these descriptions we filter out the skills they address and point to the right courses. This is feasible, we can point to MOOCs offered by different institutes and corporations, as well as blended master courses and other training opportunities.
So we can analyse emerging industry needs, we can pinpoint necessary skills, we can determine the skills gap within a resume and we can point to existing courses … [slight pause] Now all that is left for us to do, is to suggest personalized learning options…[longer pause]
[visual of me screaming loud – with live audio] Step 5 is the truly scary part. Step 1 to 4 are all about matching and straightforward algorithms. But step 5 is about learning. Wait, let me ask you a little question to reflect on… [small pause] What is learning? [reflection pause]
This is the problem with learning. It is immensely complex. We can use algorithms as far as to locate courses from the net and match them to a person’s missing skills. But as we all know there is more to learning than simply pointing to existing courses.
The skills engine, provides a cluster of useful courses coming from different institutions and organisation. But are these clusters of content useful for the learner? What might be the challenges?
There are a lot of challenges when it comes to providing useful personalized learning content to learners. Let me zoom in on three key challenges that are familiar to all of us, as we have been tackling these for years.
Granularity is a challenge. It was a problem in the past as we tried to find the right solution to update online learning content and keep it useful. With the skills engine it becomes a challenge, because if we cluster courses, we need to avoid overlapping content. And we need to make sure that the building blocks aren’t too distant from each other. We need to make sure the learner can find their way through the content cluster that is offered.
Pedagogical scaffolding is another challenge. In the past one professor was responsible for providing logical teaching steps in his or her course material. But if we cluster content which is produced at different institutions, and we reassemble these content chunks into new learning tracks, chances are very real that these content bites have a different pedagogical underpinning. So we need to find the right support.
What about certification throughout the process? It used to be that one university provided the degree. Lately we have seen that MOOC providers offer a more micro-credit-oriented certification, or open badges. So the certification is getting more blurred. But now, with the skills engine, as content originates from different sources, we need certification from all of these content providers. How will we do this? Can we offer a way to stack up learning into an up-to-date certification process? Blockchain in education is an option, if you are interested in this, feel free to join the Blockchain in Education session that is planned Friday at noon in another session.
So let’s be honest, there are a lot of obstacles on the way, but at least we know where we need to head [retake von Neumann quote]
If you list all of the challenges, and there are plenty, it becomes clear that we cannot solve all these challenges by ourselves. The good news is, we don’t need to. There is a growing set of off-the-shelf AI tools that make life easier to get this overall solution in place. We just need to assemble the right pieces.
Let me highlight just three of them that we are testing to embed in our skills engine.
Filtered http://learn.filtered.com/. Previously called Magpie, offers small open educational resources from the net like tedtalks, youtube movies around learner driven content. The AI also looks at quality of the resources and relevance to the subject matter for the learner (UK).
[visual Sproutlogix and short list others in clear visual language]: Adaptive learning experiences is another tool set that we are exploring We are looking into Sproutlogix DeveLoop - https://www.sproutlogix.com/. This adaptive learning experience platfom uses Natural Language Processing, behavioral science and data analytics to create individualized learning paths and further recommendations. It also provides a baseline pre-assessment to evaluate competency proficiency levels. Integrates with LMS. (India)
[visual wildfire learning]: And then a tool which offer opportunities when in case there is a lack of content Wildfire Learning http://www.wildfirelearning.co.uk/. Is a semantic tool that takes any document, PowerPoint or video (eg. Wikipedia, YouTube, corporate documents) and converts it to interactive (fill in the blank) questions to test knowledge. From Donald Clark who is here at the conference as well.
Now, how far have we come since that dream of a skills engine? Just imagine in just a couple of months, thanks to the skills engine, an employee will save time to find a job, as well as the necessary courses to skill her or himself for that new job and become better at it. Employers will be enabled to pro-actively train their employees for future jobs within the company. [pause] But this innovation is bigger than a simple list of benefits in a search for efficiency. We – as learners – will be able to take our future into our own hands. We will be able to find a more suitable job, a job that really fits our capacities and skills. We will all become more motivated in the process, as it fits our skills. Employers will be able to create a company where people love to work because it really suits their profiles. If we can do this, then we are making all of us happier, we surpass efficiency only, and create more humane working environments.
I hope to talk to you within this session or later during the conference. Thank you and looking forward to exchanging ideas!
Building the Skills Engine: our dreams realise the future
Our dreams realise the Future
Inge de Waard
InnoEnergy is supported by the EIT,
a body of the European Union
• Inge de Waard
Team effort, thank you: Yves Peirsman, Marloes
Wichink Kruit, Frank Gielen
& the Techwolf team: Andreas, Jeroen & Mikaël.