Virtues

2.4. Virtues#

In carrying out a principled Bayesian approach to learning from data, there are virtues that we wish to uphold. A good list of these is reproduced here from the paper by Gelman and Hennig. These are broadly relevant for the practice of science and not only Bayesian inference.

Aspirational virtues for Bayesian inference and beyond

  1. Transparency

    • Clear and unambiguous definitions of concepts

    • Open planning and following agreed protocols

    • Full communication of reasoning, procedures, spelling out of (potentially unverifiable) assumptions and potential limitations

  2. Consensus

    • Accounting for relevant knowledge and existing related work

    • Following generally accepted rules where possible and reasonable

    • Provision of rationales for consensus and unification

  3. Impartiality

    • Thorough consideration of relevant and potentially competing theories and points of view

    • Thorough consideration and if possible removal of potential biases: factors that may jeopardize consensus and the intended interpretation of results

    • Openness to criticism and exchange

  4. Correspondence to observable reality

    • Clear connection of concepts and models to observables

    • Clear conditions for reproduction, testing and falsification

  5. Awareness of multiple perspectives

  6. Awareness of context dependence

    • Recognition of dependence on specific contexts and aims

    • Honest acknowledgement of the researcher’s position, goals, experiences and subjective point of view

  7. Investigation of stability

    • Consequences of alternative decisions and assumptions that could have been made in the analysis

    • Variability and reproducibility of conclusions on new data

George Pólya was a Hungarian-American mathematician who, in addition to many contributions to pure mathematics, developed a framework for problem solving that incorporated many principles of plausible inference that we build into our Bayesian methods. He also articulated three moral qualities essential for the robust practice of scientific discovery in general, which we have adapted here with the hope they are shared by all of our readers.

Moral qualities of the scientist

Adapted from G. Polya, Induction and Analogy in Mathematics, chapter 1, section 4, which is entitled “The Inductive Attitude”.

Intellectual courage

One should be ready to revise any one of our beliefs.

Intellectual honesty

One should change a belief when there is a compelling reason to do so. To stick to a conjecture clearly contradicted by experience is dishonest. It is also dishonest to ignore information or not to state and criticize all assumptions.

Wise restraint

One should not change a belief wantonly, without some good reason. Don’t just follow fashion. Do not believe anything, but question only what is worth questioning.