Towards Accountability in Machine Learning Applications: A System-testing Approach

61 Pages Posted: 18 Feb 2021 Last revised: 12 May 2022

See all articles by Wayne Xinwei Wan

Wayne Xinwei Wan

University of Cambridge - Department of Land Economy

Thies Lindenthal

University of Cambridge

Multiple version iconThere are 2 versions of this paper

Date Written: May 12, 2022

Abstract

A rapidly expanding universe of technology-focused startups is trying to change and improve the way real estate markets operate. The undisputed predictive power of machine learning (ML) models often plays a crucial role in the 'disruption' of traditional processes. However, an accountability gap prevails: How do the models arrive at their predictions? Do they do what we hope they do—or are corners cut?

Training ML models is a software development process at heart. We suggest to follow a dedicated software testing framework and to verify that the ML model performs as intended. Illustratively, we augment two ML image classifiers with a system testing procedure based on local interpretable model-agnostic explanation (LIME) techniques. Analyzing the classifications sheds light on some of the factors that determine the behaviour of the systems.

Keywords: explainable machine learning, accountability gap, computer vision, real estate, urban studies

JEL Classification: C52, R30

Suggested Citation

Wan, Wayne Xinwei and Lindenthal, Thies, Towards Accountability in Machine Learning Applications: A System-testing Approach (May 12, 2022). Available at SSRN: https://ssrn.com/abstract=3758451 or http://dx.doi.org/10.2139/ssrn.3758451

Wayne Xinwei Wan (Contact Author)

University of Cambridge - Department of Land Economy ( email )

19 Silver Street
Cambridge, CB3 9EP
United Kingdom

HOME PAGE: http://xinweiwan.weebly.com

Thies Lindenthal

University of Cambridge ( email )

Trinity Ln
Cambridge, CB2 1TN
United Kingdom

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
70
Abstract Views
375
rank
379,080
PlumX Metrics