Risk mitigation in algorithmic accountability: The role of machine learning copies

Abstract

Machine learning plays an increasingly important role in our society and economy and is already having an impact on our daily life in many different ways. From several perspectives, machine learning is seen as the new engine of productivity and economic growth. It can increase the business efficiency and improve any decision-making process, and of course, spawn the creation of new products and services by using complex machine learning algorithms. In this scenario, the lack of actionable accountability-related guidance is potentially the single most important challenge facing the machine learning community. Machine learning systems are often composed of many parts and ingredients, mixing third party components or software-as-a-service APIs, among others. In this paper we study the role of copies for risk mitigation in such machine learning systems. Formally, a copy can be regarded as an approximated projection operator of a model into a target model hypothesis set. Under the conceptual framework of actionable accountability, we explore the use of copies as a viable alternative in circumstances where models cannot be re-trained, nor enhanced by means of a wrapper. We use a real residential mortgage default dataset as a use case to illustrate the feasibility of this approach.

Application of Optimized Support Vector Machine Model in Tax Forecasting System

“…In his research, Hao et al proposed an algorithm for financial risk prevention and carried out special data preprocessing on convolutional neural network. Combined with the requirements of digital inclusive financial risk method, he constructed a digital inclusive financial risk prevention model [13][14][15], so as to timely find financial abnormalities and carry out risk early warning.…”

Differential Replication for Credit Scoring in Regulated Environments

“…Due to proprietary reasons, the original dataset cannot be made public. A reconstruction of this dataset that recovers the mean, standard deviation, covariance, and correlations, as well as the label distribution can be found in [ 4 ].…”

“…This context includes dimensions which are both internal and external to the organization and which are mostly out of our control [ 2 , 3 ]. More worryingly, these are all dimensions that refer to constraints prone to change in time [ 4 ].…”

The Challenges of Machine Learning and Their Economic Implications

“…The need to align with these business objectives is just one of the numerous factors that constrain the commercial deployment of machine learning. When used in any company or public institution, machine learning models are constrained by the data they are trained with, as well as by the governance model that controls these data, the different stakeholders that interact with them throughout their lifespan, the software licenses, the regulatory framework, the need to preserve industrial secrecy, and, ultimately, the technological infrastructure that serves the models into production [13]. These factors collectively form the ecosystem of a model and can be understood as constraints that limit the form and the format of the deployed solutions.…”

“…Nevertheless, the environment in which machine learning models are trained and deployed is usually complex and introduces several constraints on both the form and the format of the resulting systems [11][12][13]. As a consequence, the increasing use of machine learning has raised concerns regarding, among others, its interpretability [14], fairness [15], safety [16], and privacy [17][18][19].…”