Summary

"Blindness" by José Saramago presents a model of interpretive blindness (IB), an epistemic bias hindering learning from testimony. IB arises from a co-dependence between background beliefs and interpretation, amplified by contemporary testimony's argumentative completeness. This bias prevents agents from updating beliefs with new, incompatible data, as they discount challenging evidence and reinforce trust in existing, limited sources. The book argues that this dynamic process, where belief updates are reinforced by the interpretation of testimony, can lead to a self-perpetuating cycle of biased learning, even within sophisticated Bayesian frameworks designed to promote good epistemic practices.

The core argument is that individuals can become "interpretively blind" when their existing beliefs shape how they interpret incoming information, particularly from testimony. This leads to a discounting of contradictory evidence and a strengthened reliance on trusted, but potentially narrow, information sources. Readers learn about a specific cognitive bias that can lead to a rational persistence in flawed belief systems, even when faced with opportunities for learning. The book demonstrates how argumentative completeness in testimony, where sources can counter any doubt, exacerbates this issue, making it difficult to dislodge ingrained biases.

Key concepts

  • Interpretive blindness (IB)A bias preventing learning from testimony due to a co-dependence between background beliefs and interpretation.
  • Argumentative completenessA characteristic of testimony where authors can respond to and argue against doubts, reinforcing credibility.
  • Bayesian settingA mathematical framework for reasoning about probabilities and belief updates.
  • Hierarchical Bayesian settingsAn extension of Bayesian frameworks used for more complex belief structures.
  • Discount and [co-dependence]A situation where agents discount evidence incompatible with their beliefs due to a co-dependence with their sources.

From the book

IB is a special problem for learning from testimony, in which one acquires information only from text or conversation. We show that IB follows from a co-dependence between background beliefs and interpretation in a Bayesian setting and the nature of contemporary testimony. We argue that a particular characteristic contemporary testimony, argumentative completeness , can preclude learning in hierarchical Bayesian settings, even in the presence of constraints that are designed to promote good epistemic practices. 1 INTRODUCTION In this paper, we describe and analyze an as far as we know theoretically un-examined kind of bias, which we call interpretive blindness (IB). IB is exemplified by humans (and perhaps soon by sophisticated machine learning algorithms) whose beliefs are guided and…

Popular questions readers ask