Proxy Optimization and Purpose Displacement: Why Paying for the Count Can Change What Is Counted

This conceptual paper introduces Proxy Optimization and Purpose Displacement, a mechanism in the Representation and Measurement series of On Human Modeling. When a reward depends on a measure, people have a reason to change how they produce the count. Some ways of raising it advance the purpose it was chosen to indicate. Others earn the same reward while leaving that purpose unserved. The paper asks what conduct raised the count and what independent evidence shows whether the intended result followed. In a rat-control campaign that began in 1902, Hanoi’s colonial authorities offered one cent for every rat tail brought to the municipal offices. A tail was easier to handle than a carcass and was accepted as evidence of a kill. When live rats appeared in the city without their tails, observers had evidence that a hunter could collect the bounty and leave the rat alive. The campaign’s overall effect on the rat population remains open. Wells Fargo tied incentive pay to sales targets for additional products sold to existing customers. In a review covering May 2011 to July 2015, the bank estimated that employees may have opened about 1.5 million deposit accounts and submitted about 565,000 credit card applications without customer authorization. The Consumer Financial Protection Bureau linked the practices to sales targets and bonuses. An account opening could count toward an employee’s target even when the customer had never requested the account. The paper places the mechanism within Donald Campbell’s account of how the use of indicators in social decisions can distort the activity they measure. A similar question arises if a team earns a bonus for increasing the share of an AI assistant’s answers that include a citation. Adding a citation to an unsupported answer raises the score. Reading the cited source to see whether it supports the answer tests the purpose for which the citation was requested.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24408931
Primary Topic
Accounting and Organizational Management
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Proxy Optimization and Purpose Displacement: Why Paying for the Count Can Change What Is Counted

Alexandra Cary
Zenodo (CERN European Organization for Nuclear Research)
Accounting and Organizational Management
article

Proxy Optimization and Purpose Displacement: Why Paying for the Count Can Change What Is Counted

Alexandra Cary
article en

Abstract

This conceptual paper introduces Proxy Optimization and Purpose Displacement, a mechanism in the Representation and Measurement series of On Human Modeling. When a reward depends on a measure, people have a reason to change how they produce the count. Some ways of raising it advance the purpose it was chosen to indicate. Others earn the same reward while leaving that purpose unserved. The paper asks what conduct raised the count and what independent evidence shows whether the intended result followed. In a rat-control campaign that began in 1902, Hanoi’s colonial authorities offered one cent for every rat tail brought to the municipal offices. A tail was easier to handle than a carcass and was accepted as evidence of a kill. When live rats appeared in the city without their tails, observers had evidence that a hunter could collect the bounty and leave the rat alive. The campaign’s overall effect on the rat population remains open. Wells Fargo tied incentive pay to sales targets for additional products sold to existing customers. In a review covering May 2011 to July 2015, the bank estimated that employees may have opened about 1.5 million deposit accounts and submitted about 565,000 credit card applications without customer authorization. The Consumer Financial Protection Bureau linked the practices to sales targets and bonuses. An account opening could count toward an employee’s target even when the customer had never requested the account. The paper places the mechanism within Donald Campbell’s account of how the use of indicators in social decisions can distort the activity they measure. A similar question arises if a team earns a bonus for increasing the share of an AI assistant’s answers that include a citation. Adding a citation to an unsupported answer raises the score. Reading the cited source to see whether it supports the answer tests the purpose for which the citation was requested.

Zenodo (CERN European Organization for Nuclear Research)
Accounting and Organizational Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon