Algorithms help us turn complex stories into dara that can drive a decision. Nearly every grading system and student performance tracker education has is based on algorithms, which is why, perhaps, they feel so useless to learners and teachers when, in some places, that is all they are supposed to care about.
This Wired article (http://www.wired.com/2013/06/ups-astronomical-math) talks about delivery firm UPS and its new software for answering "the traveling salesman's problem", namely: how to get from A-Z via all the required points in the middle most efficiently. The software is not algorithmic, that is, it's not written in a lab, shipped and then "done to" the driver. Instead, the driver can use his or her often superior knowledge of local quirks in traffic at certain times of day, those lanes and side streets that shave minutes of a journey, even if it is not as a crow would fly. The software then learns from this deviation and helps all UPS drivers do better next time.
It is heuristic, not algorithmic, and this is interesting for tracking software of any kind because heuristics, while always possible in maths and programming, was hard to do better than a human until relatively recently.
What would this mean for learning analytics, a host of services and products more often arousing ideas of big money than big student progress? It means that the decisions of the most successful students could land data that helps weaker students perform better. It means that the subtle pedagogies of self-reported grading and quality feedback might begin to have a hope of being undertaken, and undertaken WELL, by software.
That would be the kind of learning analytics I might start to get excited by.
Recent Comments