Page 349 - Deep Learning

332 Conversion

Neither the everyday practice of using the terms “knowledge” and “belief”
to designate degrees of certainty, nor the normative definition of knowledge
as justified true belief will play any role in this book. I most often use the term
“knowledge” in the cognitive psychology usage, as encompassing a person’s
beliefs about a topic. Occasionally, I use it in its overarching, cognitive science
sense of referring to all mental representations, including nonpropositional
representations (e.g., visuals images). The context should suffice to make clear
which of the two meanings is intended in each case.
The truth value and the confidence level associated with a belief are distinct
from the person’s attitude toward the belief. Does he view the state of affairs
described by the relevant proposition as desirable or as undesirable? Attitudes
are, in principle, orthogonal to truth values. Disbelief is not the same as dislike.
For example, a stamp collector might believe that the world is switching from
physical to electronic mail, and evaluate this state of affairs negatively because
there will soon be no more stamps to collect. Another, shrewder stamp collec-
tor might have exactly the same belief about which way the world is going but
see the development as positive, because he believes that his stamp collection
will increase in value as stamps become scarce. The owner of the store who
sells stamps to the two collectors might disbelieve that stamps are disappear-
ing, which is a different state of mind from that of the other two.
The role of attitudes in belief formation has received considerable atten-
tion from psychologists because common sense suggests and systematic
3, 4
studies confirm that attitudes and subjective truth values are correlated. We
tend to accept as true propositions to which we have a positive attitude and
doubt propositions that describe states of affairs that we see as negative. In
everyday life we call this wishful thinking; in psychology it is called belief bias.
But we sometimes accept uncomfortable propositions as true, and we some-
times decide that some claim is too good to be true, so the correlation is not
perfect; we cannot equate judgments of truth and judgments of desirability.
The problem of attitude change – how and why do people come to like what
they previously disliked and vice versa – has been studied extensively by social
psychologists. The theory proposed in this chapter is not intended to contrib-
ute to that field but to clarify the separate problem of how and why people
come to believe what they previously disbelieved or vice versa.*

* Confusingly, some philosophers refer to facts like “Mary believes that the Earth is flat”
as propositional attitudes. (Moltmann, 2003, describes the classical view of propositional
attitudes in analytical philosophy, critiques it and proposes an alternative.) This usage of
“attitude” differs from its psychological use and will be avoided here.