Big data algorithms can help detect correlations. The software does the job much faster than humans.
MIT researchers aim to take the human element out of big data analysis.
Their approach seams good in finding patterns in big data that data analyst should regard first.
Therefore it is a great asset to help data analysts to focus.
Correlation does not imply causation. Causation is absolutely important for any decision process. For causation we need human data analysts. But there is too much unanalyzed data available for the few data analysts in the world. Can we digitalize them?
We are far away, although we know some good approaches: Rule-based expert systems, fuzzy logic systems and fuzzy cognitive maps can model human reasoning. If you want to save data analysts later in the processing of big data, you will need them before for the deduction of the rules needed.
Besides the human ability to derive rules for our model, data analysts have another secret weapon. It is intuition. This intuition is build also on rules and can be regarded as biased. Even for human it is not easy to perceive and write down these kind of rules. Assumed that we can extract them, shall we really implement them as artificial intuition? It will give us the chance for quick decision making in uncertainty.
But those rules can be also be frightened, right?
Do you fear biased machine learning?
If you want to reduce implicit bias from algorithms, we can use artificial methods also.
Again you will need rules to discover biased algorithms.
In summary this will lead too many rules and layers of rules. Nobody will have the time to maintain them.
May be a human data analyst is not a bad choice. What do you think?
Keine Kommentare:
Kommentar veröffentlichen