A Quick Post on Deduction

As we discovered in a previous post, inductive reasoning has its flaws. Mostly these arise because we are vulnerable to false positives – to causal links that seem to be real but in actuality are not there. We are prone to mistaking noise for signal.

Deductive reasoning, the alternative approach, does not rely on observational reasoning and, hence, avoids this error. Deduction reasons from a set of axioms, using logic to arrive at conclusions. One identifies rules determining the behaviour of, and interaction between, elements of a system, or the system as a whole, and arrives at conclusions by following a logical string of arguments. Sounds like fun.

Imagine you are deciding whether to go to a (probably disappointing) house party or not. You could use deduction by identifying that you like hanging out with the people going, you like drinking alcohol, you like music, etc. You deduce that if add these elements into one experience, you will enjoy that experience. To argue inductively would be to recognise that you have enjoyed similar events (other mediocre house parties) in the past and use this to guess that you will probably enjoy this event.

Flaws

Deduction is not a panacea. For starters, the method is reliant on the axioms and logic applied to those axioms. If either are wrong, your conclusions will be wrong too. Deductive arguments tend to be slightly more theoretical and ideological. Induction moves from reality to theory whereas deduction moves in the opposite direction. At least when one uses induction, one knows the reality part is definitely correct. A deduction may look cute, but it might not be a reflection of reality.

Deduction (green) vs. induction (red)

Ultimately, both methods require testing and experimentation. Both must make testable predictions, or risk being trapped in the realm of nice-sounding theory that no one can prove and is therefore pretty useless. The main problem is that in many domains making testable predictions is difficult. I repeat, life is not a laboratory. It’s difficult to conduct good experiments and so difficult to prove/disprove theories, either deductive or inductive. What we are left with in most cases is a tangled mess of anecdotes, theories, faulty axioms, logic, correlation, mistaken causation, fake data, changing statistical relationships, wacky professors, rich-people-turned-philosophers, real scientists, perverse incentives, bad research, and…good science. All mixed in together. It’s my/your/our job to untangle this mess.


Subscribe

See Privacy Policy here.