adversarial data mining

  • thinking on how hyper-precision l, huge levels of data capture and hard mining it is dangerous in the wrong hands
  • eg segmenting a group to target, discriminating against a group by associated characteristics indirectly
  • correlation != causation
  • hyper-precise data capture with no specific goal (eg with machine learning) could surface correlations before mainstream science has this data
  • seeded by considering not everything should be IoT capable on seeing IoT toothbrushes and hypothesising a danger, what if the statistical likelihood of low dental hygiene in certain medical conditions could cause unwarranted and nonconsensual segmenting of this population of toothbrush users