Evaluation and Influence through Selective Learning of Attributes

Attributes: Selective Learning and Influence

“…Appendices C and D collect proofs and auxiliary results for Sections 3.2 and 4.2, respectively. Appendix E supports the discussion in Section 5 (see Bardhi (2024) for Appendices C, D, and E).…”

“…Appendices C and D collect proofs and auxiliary results for Sections 3.2 and 4.2, respectively. Appendix E supports the discussion in Section 5 (seeBardhi (2024) for Appendices C, D, and E).6 Given a probability space ( F P), a stochastic process f := {f (a ω)} a∈A ω∈ with index set A is a Gaussian process if and only if (f (a 1 ) f (a n )) are jointly Gaussian for any a 1 a n ∈ A and n ≥ 1. See Appendix A andRasmussen and Williams (2006) for a technical introduction to Gaussian processes.7 Appendix A provides sufficient conditions on (μ σ) to guarantee sample-path continuity (Proposition 13) and establishes that sample-path continuity implies continuity of μ and σ (Proposition 14).…”

“…The Gaussian structure is consistent with the principle of maximum ignorance: given (μ σ), the players adopt the least informative (entropy-maximizing) prior distribution over attribute realizations, which is a multivariate Gaussian.35 An earlier version of this paperBardhi (2018) considered such a binary decision setup. See Appendix E.2.…”

“…3333 In Callander and Clark (2017), optimal case selection does not consist of the most uncertain legal case. Similarly, the most uncertain attribute is not the most informative in Section 3; see Example E.5 in Bardhi (2024). …”

“…3535 An earlier version of this paper Bardhi (2018) considered such a binary decision setup. See Appendix E.2. …”