Instilling Values In Machine Learning | Avast

Byron Acohido, 4 January 2021

UK exam debacle, US pollsters’ inaccuracy show algorithms bereft of knowledge

Plato once sagely observed, “A good decision is based on knowledge and not on numbers.” 
That advice resonates today, even as we deepen our reliance on number crunching — in the form of the unceasing machine learning algorithms whirring away in the background of our lives, setting in motion many of the routine decisions each of us make daily. 
However, as Plato seemingly foresaw, the underlying algorithms we’ve come to rely on are only as good as the human knowledge they spring from. And sometimes the knowledge transfer from humans to math formulas falls well short.
Last  August, an attempt by the UK government to use machine learning to conjure and dispense final exam grades to quarantined high-schoolers proved to be a disastrous failure. Instead of keeping things operable in the midst of a global pandemic, the UK officials ended up exposing the deep systemic bias of the UK’s education systems, in a glaring way. 
Then, in November, the algorithms pollsters invoked to predict the outcome of the 2020 U.S. presidential election proved drastically wrong — again, even after the pollsters had poured their knowledge into improving their predictive algorithms after the 2016 elections.  
Algorithms reinforce UK education biases 
These two high-profile knowledge-transfer failures are instructive. They come at a time when companies are grappling with something marketing experts refer to as the “personalization vs. privacy paradox.”  
This refers to the difficulty of crunching data to influence human behaviors, without crossing a very thin line into privacy invasion. Data gathe ..

Support the originator by clicking the read the rest link below.