To infer is to work from axioms or claims to find what is implied by them. The claims may be statistical correlations.

There is no standard of evidence in inference, as it uses whatever evidence is provided to draw conclusions axiomatically, to different standards of rigor:
  • axiomatic proof which is generally thought to be very reliable but narrow - used mostly in mathematics
  • quasi-empirical methods? including highly trusted human arbitrators, such as in a very-high-stakes forensic? investigation, e.g. forensic audit?s; these will rely on reports from others and so this is only a high standard to the degree that that evidence/source/authority is accurate.
  • statistical proofs? usually measured in "sigmas" beyond the mean, that mean signalling coincidence, and basis of advanced voting systems, e.g. Single Transferable Vote, and standards of circumstantial evidence? - not hearsay? which depends on trust in a source not on any correlation.
  • statistical significance? indicating some level of correlation?
  • balance of probabilities?, colloquially, "51%", basis of civil law?, simplest voting systems.

The idea of rigor is specific to inference and should not ever be applied to any deference (to which morality applies instead) or reference (to which integrity applies instead).

It is possible to structure inference as dialectic or tensegrity but these structures are themselves composed of many inferences or balance of powers? or balance of tensions? or balance of probabilities?.

Show php error messages