4 ms·

BP oil rig modelling software showed cement unstable days before blast

• z92 16y ago
  It's rate of false positive that really matters. When that software says something is wrong at magnitude-X and with a very low false positive rate that thing is really wrong at that scale, then I shall agree it was BPs fault for ignoring it.
• jared314 16y ago
  I like the subtitle better: Oil giant decided data was unreliable and took other steps. Just remember this when the reverse happens, where the model is wrong and they ignore all evidence to the contrary.
  • scott_s 16y ago
    But did they decide the data was unreliable because they didn't like the conclusion, or because they had genuine reservations about the validity of the model and/or data? And if they did, why check the model, anyway? What worries me about the mentality of "I'm not confident in this data, but let's check the model anyway" is that being irrational creatures, we're likely to use a positive result from the model as evidence that everything is okay, but ignore a negative result because the data is unreliable. That is, different results, same conclusion. This is not logically valid, but we're irrational creatures and it takes discipline to avoid the fallacy.
    • Retric 16y ago
      I assume they did not like the conclusion. The company conducted a separate negative pressure test... The test was failed, but was – for an unexplained reason – deemed a “complete success” by both BP and rig owner Transocean at the time, a presentation on Monday said.