Skills: currency or recency?

Skills: Currency or Recency?

Skills currency has long been taken to denote a unit of value like a sovereign fiat currency in the economic analogy, but it has another meaning, currency as time. In UKGs latest blog, “Skills Recency Comes to Workforce Management Systems” they seem to be nuancing this as a proprietary mechanism in their AI/ML/NLP as a predictor of skills intelligence driving their UKG Skills Recency Index? .

Not all skills degrade

UKGs most recent blog on skills they state that "skills will inevitably degrade over time".

Many skills are on a continuum of growth over time, especially those that we now recognize as the most durable and essential skills. Those skills are the ones that drive the learning and application of new hard skills in context. And even hard technical skills persist as foundational to the skills that follow.

Undoubtedly, the rate that new skills need to be acquired has accelerated, but that does necessarily mean the prior skills are less valuable. They are often incremental on that prior stable foundation and understanding. Determining on declining skill count declarations that a skill is less valuable leads to potentially unintended decisions. They provide a compound layering effect and the basis of experience and context. Rare are the quantum inventions that do not leverage any previous knowledge or skills. We have to learn the rules of the road before jumping in the newest car.

The Tech industry has been particularly central in the perception of degrading or devaluing skills based on the acceleration of versioning, smaller iterative scope, and agile software development. These versions tend to be smaller, and more incremental and not net new skills. One version of Windows to the next is not a total reskilling exercise, yet they tend to get listed, extracted and matched separately as discrete skills. Using machine learning and NLP we can understand that they may be proxies i.e. that the current and prior version of AWS security are incremental. We can also learn that both are a proxy of ?“cloud infrastructure security” which could also be portable as an adjacent skillset to Azure or other cloud security contexts; you may need to learn the menu location and config parameters but is predicated on understanding cloud security.

?The step-change or complete paradigm shift does not occur very often. ?

The half-life skills myth

The half-life of skills statistic is often quoted and rarely if ever attributed to data. While directionally it may feel right, the data jury is out or yet to be convened.

The earliest origin I have found is an academic paper by Cathy Gonzalez in 2004. She used the exploding growth of content on the web, which she called knowledge, and a version of Moore’s law, to suggest that as knowledge volume being content on the web doubled over 10 years, then the half-life of what was known at the outset is only half of what is known now and therefore less valuable.

From there it has been repeated loudly and often as truth. Even the linked IBM paper in the blog, does not cite research or evidence to where they quote it from; it’s stated as fact.

Context vs Time as Capability or relevancy?

Time is not a great proxy for proficiency, and capability, let alone relevance. Its something that machine learning can look for as low-hanging fruit in the search to make sense of the masses of chronologically organized work data in resumes and profiles.

When a job req says “10 years experience”, what does that mean, time served, tenure in role? That’s not where skills is taking us. It’s especially true when the JD has the “all other duties as required which is really where the outcomes happen. The responsibilities of a JD are rarely written with a time constraint in them. Goals and KPIs are but they are rarely part of the training data to train machine learning. The time piece is usually in the requirements and skills section, but for what to justify the title, seniority, pay level,…? Is it really relevant?

Take IT Disaster recovery, or Forensic Cyber Security; not necessarily skills you want someone to be having to use every day as that that would indicate some significant other issues. But you do need to know who has them in your inventory. Of course, you can run simulations to maintain currency. In certain professions and industries, it is and has to be done well for regulation, and compliance, but across the vast majority?

Time over skills occurrence in roles can also be misleading for experience as perhaps year 1 of experience is merely repeated over many years; time does not denote a growth mindset of increasing or broadening contexts, expertise, and application. Sophisticated linguistics can help us look for words and grammar that do indicate more of that progress. ?

Time based skill prevalence in a role profile or description may also severely skew prediction of ramp time. An individual may take a different choice about sharing as explicit or internal learning journey depending on motivation.

We don’t all publish our diet or gym goals, not everyone publishes their career aspirations or vulnerabilities.

Skills is not data science, it is the science of data.

This is a challenge with many Skills Intelligence solutions in the data they have based on what people choose to publish in an enterprise tool that’s HR done to them, and that that’s inside the current employer of record firewall data, to determine learning activity as its very murky in the labour market for prior roles. While the externally crawled labour market data can be applied that’s based on profiles that tends to be limited as few profiles include those internal LMS learning journeys on the anonymized 2 page resume.

The 5 use cases listed in the blog seem more like a Turing test for whether relevancy is a fair proxy that customers should be using to audit the machine learning of its potential vendors.

Skills is not about the simple elemental identification of single nouns, Likert scales and axes, but about their properties too, their observable demonstrable outcomes to enable people to understand where they are, the non-linear curve of their progress over time, their bonding impact when paired or combined with other skills that an individual may have, or that produced results when paired with a prior colleague or teammate(s).

Evidence and context are critical in the understanding of capability. Skills is an applied science of experimentation, and results or outcomes observation; what combination of skills in the compound context produce what outcomes? These tend to be more readable in learning outcomes, credentials, OKRs, and KPIs of performance demonstrated and achieved, and more subtly in mentoring, coaching and management feedback but we’re very poor about capturing and curating those discussions and progress as evidence of current capability.

While time waits for no one, its steady progress does not normalize for the currency of people.