Skip to content

ACCA Strategic Professional · Advanced Performance Management · Data science and analytics

Harbor Logistics uses an algorithm trained on five years of historical manager ratings to recommend annual pay rises. Analysis shows female drivers receive systematically lower recommendations for equal measured output. What is the most likely root cause and best first response?

The likely cause is bias embedded in historical manager ratings used as training data, which the algorithm reproduces. The best first response is to review and rebalance the data and test outputs for fairness before relying on it. Hardware, random deletion or charts do not remove bias.

  1. ABias embedded in the historical training data; review and rebalance the data and test outputs for fairness before useCorrect
  2. BInsufficient computing power; upgrade the servers
  3. CExcessive data volume; delete half of the records at random
  4. DPoor data visualisation; change the chart format

Explanation

A model trained on past subjective ratings learns and repeats any bias in them. The appropriate response is to examine and correct the training data and test outcomes for fairness. Computing power, random deletion and chart format do not address biased labels.

Did you get it right without looking?

One question tells you little. A timed set on Data science and analytics shows your real accuracy, how long you take and where you lose marks.

Practise Data science and analyticsFree diagnostic mock

More Data science and analytics questions