> I do think we should use citizen data to schedule shifts though. It makes no sense to me, to have 10 nurses and 15 teachers do full time scheduling in a city of 60,000-100,000 citizens when an algorithm can do it instead. But we can’t, because the GDPR prevents us from using data that way.
Making sure the right staff gets to the right citizens at the right time is a massive undertaking. It’s where 60-80% of our staff works. We have something like 1200 teachers out of 7000 employees for instance.
It’s also an area that’s subject to a lot of requires/shared resources. I mean, you have the regular schedule which takes up a lot of time on its own because you need to find replacements when someone is missing. If I get sick, no one really cares that we go a day without an Enterprise Architect, but if a teacher goes sick, multiple classes will need an replacement. Then there is the irregular stuff like rehabilitation, special school events or a range of things.
Basically it’s so complicated that it takes up the time of several full time positions.
It’s not the kind of complicated that’s not solvable by ML though. Our neighbouring municipality did a PoC on it, and I’m not sure how the procured the rights use the data for it (they probably didn’t), and it turned out that they could automate almost all of the planning and scheduling. Humans still had to make decisions, but it would suggest available resources or alternative schedules making the process much smoother.
That’s a lot of nurses and teachers you could actually put to work, nursing or teaching if it works.
It works by knowing what resources are needed though, and in the case of rehabilitation that involves knowing that someone needs anti-suicidal therapy at 10 am at a certain address. Which is extremely sensitive data, that the algorithm isn’t allowed to access under our adoption of the GDPR.
Under the implementation of GDPR, or under derived local rules? Because as far as I understand, humans working on a filing cabinet full of paper can be just as much "data processing" under GDPR as a computer reading a database. If the scheduling person knows "Mr X is sick and needs a replacement scheduled", an algorithm in their place can know that too.
It's certainly something where extra care is required though, and it's easy to see how people in charge go with a rather safe than sorry approach, and extending it into ML poses additional questions.
Our national implementation of the GDPR for the public sector prohibits cross-sectoral access to citizen data.
So while our teachers/nurses could legally use ML for planning as they have a legal right to use the data for planning, our digitisation department can’t build/train/support it and neither can a 3rd party supplier.
Since laws are very open to interpretation, at least until they are tested a few times in the courts, you could interpret them different than us. Which I’m guessing is what our neighbouring municipality is doing. They have the advantage of being 10 times bigger than us though, giving them much more influence, so much in fact, that they may end up paving the way for the rest of us.
Can you give more details on that case?