Occam’s Razor – cutting edge or dulled after 700 years?

The phrase often associated with William of Occam (or Ockham) “Entia non sunt multiplicanda praeter necessitatem” (Hypothesis should not be multiplied without necessity) may not actually be the words of William of Occam but certainly Occam wrote of a similar concept with “Numquam ponenda est pluralitas sine necessitate” (Plurality should not be postulated without necessity) and the sentiment certainly stands in both cases.? In essence there are two interpretations of the philosophical razor used.?

·??????? One should not go further than evidence allows.?

·??????? Seek the solutions which involve the least steps to achieve

These concepts have been handed down through many scientists thought the centuries, but in the modern age of big data and high-profile cases do these simple rules still play out?? Has Ai and data science moved to such a degree that we can take a more nuances approach or is this the acceptable face of speculation?

Making claims

When we make a claim in a safety case (or indeed cyber security or any domain) the claim represents to the best of our knowledge the capability of the system.? The claim should therefore be complete, accurate, understandable, and constrained.

Looking at Occam’s razor we can elaborate on the four properties.

• Complete – The case should contain all the system properties, behaviours and claims which can be directly measured.? Those which required derivation or are implied should not be included or the basis of the derivation clearly defined.
• Accurate – The case should contain unbiased information which can be substantiated directly.? The inclusion of information derived or implied can be dangerous as opinions and bias can become fact.
• Understandable – the information should be directly stated and not require the reader to make multiple steps to gain useful understanding.? This ?can lead to incorrect interpretations being made.
• Constrained – How was the evidence derived?? What was included in scope.? These questions constrain the case and the behaviours.? This ?should be clearly stated to ensure that the correct context is given.

Data Entropy

Data entropy as a concept looks at the degree to which a dataset has become disordered over time and with manipulation.? Within our discussion this concept is important as with each process which is applied to data a small amount of the data is filtered out.? These filters whist designed to bring order to data may lose or slew the meaning of the data to that which act to confirm the bias of the person which has applied the filter.? By this end inconclusive raw data is used to “prove” a postulated argument.

Therefore, to made true data driven decisions Occam requires that the filtering is kept to a minimum which is required for processing or clarity and that that filtering is not giving a false correlations.

When processing data for claims, it is important that the number of processes is assessed.? This gives a clear picture of the degree to which the information has been filtered.? Ideally each piece of information is directly measurable, however this is? not always the case and a degree to? analytics is required.? Where this is the case the steps taken and assumptions made along the way should be included to ensure that a complete case can be given.

Conclusion.

Is Occam’s razor dulled? No, it is as sharp as ever.? The only essential in this is to use it.