Thursday morning Keynote: Educations vs Training

Selena Deckelmann (@selenamarie) gave the Thursday keynote address at LISA this morning. Her topic: Educations vs Training. We’ve had a Workshop on this earlier, and a BoF session last night so we’ve been talking about it all week.
 
Unfortunately, we’ve been talking about it here at LISA since 1997 and the fundamentals of that talk haven’t changed. This is bad, and we need to stop that and actually change how we’re talking about it. The problem is only getting worse; the Bureau of Labor Statistics predicts that by 2018 there will be 1.5 million additional jobs in computer-centric careers like ours, and we’re having trouble finding decent workers right now.

The mindset we’re stuck in right now is that Education does not equal Training. Education will teach you how design a wiring diagram, know how to get at the management interfaces of networking gear, and similar. Training tells you that talking down a specific network switch will bring the HPC cluster down, or that HSRP routers won’t faillback a Windows service because Windows won’t issue a Gratuitous-ARP packet without a specific hotfix. Because of this, there is a wide suspicion that degree programs at formal educational institutions are largely pointless since they simply can’t handle the training aspect, and that is where most of systems administration domain knowledge resides.

Selena challenges this. The training and education domains actually overlap, and overlap quite a lot. The overlap is where knowledge and skills are centered. Education provides habits, training provides competency. It all works together.

Bloom’s Taxonomy describes the orders of thinking. At the bottom is Remembering, rote memorization, moving up through understanding, applying, analyzing, evaluating, and finally at the apex: creation. Education is really good at providing the first three steps, training is good at the top three.

Selena has been working withPyLadies to bring more women into the Python community. One of the ways they do this is to provide basic programming training. The women coming to Python probably didn’t have any programming courses, and may be missing even more fundamental background knowledge like algebra (it’s hard to grasp variables if you did badly in algebra). They’ve built their educational process to handle all of this.

They start with the very basics, the functional stuff of how do you write this at all (remembering and understanding). Then move into more complex areas, like variables, loops, flow-control, and data-structures (applying). Once these are mastered they write exercises (applying & analyzing). Then they go through a section on debugging other people’s code (evaluating).

This is a very hands on process and involves a few other techniques to keep students engaged. There are defined outcomes. Instruction comes with feedback loops for both the student and mentors. The whole process is designed around collaboration, which further trains by modelling and pairing. At the end students are able to read code, even code much more complex than they could produce themselves.

This training model can be simplified to four characteristics: numbered steps with measurable, defined outcomes; explicit instruction and feedback loops; expectation for collaborating; pairing and modelling. Selena’s model works, but we need more voices in the conversation. Selena advocates picking constructive fights. There is not currently enough discussion between educators and practitioners to share needs and experiences. We need to discuss a few key questions: is it more important to initially focus on ethics or on risk management? Is K-12 the best place to focus on education, or does it belong at the Master’s level? How do we create a non-profit for supporting trusted certifications?

One questioner has been doing a lot of this already, but it is very expensive with a 2 or 3 to 1 student:teacher ratio. Universities can’t scale to this. Selena pointed out that once professional educators get involved this should improve. Questioner also suggested that getting practitioners in to do a bit of mentoring it’ll help a lot as well.

Another questioner pointed out  that current sysadmins come up through the ranks and get diverse training along the way. This creates a very diverse skill base. Would moving to a more formalized system reduce that kind of diversity? Selena pointed out that by creating a formal system that is a lot more approachable, and is recognized externally as valid, it’ll increase our overall diversity as a result.

Another questioner points out that the big problem is thinking, not technology. An earlier questioner from NASA pointed out they have a good amount of really old systems (WinNT! PDP11’s!)  that are still in production for very good reasons, and worried about getting talent in a more formalized environment. This questioner made the point that per-organization training is still going to be needed to cover the specific use-cases for that organization; formally trained sysadmins will have the thinking skills, they just need an update on the details of the new environment.

There are a lot of opinions on this, and they need to be brought forward to disrupt, and hopefully improve, the conversation around making new systems administrators.