AI predicting risky choices

Black Box Breakthrough: Can AI Predict Your Risky Choices?

Elliot Brynn in Business & Economy September 2025 • 4 min read.

"Unlock the secrets of economic anomalies with AI-driven predictive algorithms – a new frontier in understanding decision-making under risk."

How do we make economic theories better? One way is an understanding of patterns. The “Allais paradox,” a classic example, questioned whether expected utility theory matched real-world choices. Economist Maurice Allais created menus of lotteries that exposed contradictions in the theory, spurring new ways of thinking about risk. In this paper, we explore how machine learning helps uncover these anomalies.

Anomalies aren't just old news. They consistently improve how we understand economic behavior, even in areas like asset pricing, game theory, and intertemporal choice. The traditional approach involves constructing anomalies, rigorously testing them, and then developing new theories to explain them. But what if machine learning algorithms could speed up this process?

Machine learning could be a tool in accelerating the development of economic theory by revealing predictive signals. The problem is that these algorithms are ‘black boxes.’ It is hard to really determine what machine learning has discovered, which are often buried in the opacity of complex function classes. That's why we need ways to automatically construct anomalies from these predictive algorithms.

Decoding the Black Box: How Algorithms Generate Anomalies

The procedures in this paper outline output anomalies that economists can scrutinize, much like the Allais paradox. In these procedures, we use supervised machine learning to create a predictive model, serving as our empirical intuition. We then contrast this model with existing theory, pinpointing the minimal instances where theory fails to align with the algorithm's predictions. These algorithmically generated anomalies become starting points to understand inconsistencies.

To set up a framework, we focus on scenarios with input features (x) and modeled outcomes (y); for example, x represents a menu of lotteries and y represents the choice probability. Theories are modeled as mappings that return relationships. These correspondences summarize the theory’s implications: for any x, the correspondence specifies acceptable y values. Expected utility theory suggests individuals make choices based on what they've already chosen in similar scenarios. So, within this framework, we define anomalies as the smallest set of examples inconsistent with the theory.

Compatibility: Theory must be compatible or incompatible with any data.
Consistency: Compatible data must align with theoretical predictions.
Refinement: Theory's implications sharpen with more data.
Non-Trivial Implications: There must be data where theory provides meaningful predictions.

To generate anomalies from a prediction function, anomaly searching is framed as an adversarial game between a falsifier and the theory. The falsifier proposes features, and the estimated prediction function assesses these features, the theory then tries to explain them by fitting an allowable function. The falsifier seeks collections where the theory struggles, leading to anomalies. In simple terms, these are scenarios the theory can't explain.

The Future of Economic Discovery

Our goal wasn't to simply revisit risk and expected utility theory. Instead, this example highlights the broader potential of anomaly generation procedures. Their success in identifying new anomalies in a well-studied area suggests they could be valuable elsewhere. Supervised machine learning algorithms can expose novel empirical patterns that existing theories miss. Rather than a black box, these procedures highlight small collections of examples to help researchers refine existing theories.

About this Article -

This article was crafted using a human-AI hybrid and collaborative approach. AI assisted our team with initial drafting, research insights, identifying key questions, and image generation. Our human editors guided topic selection, defined the angle, structured the content, ensured factual accuracy and relevance, refined the tone, and conducted thorough editing to deliver helpful, high-quality information.See our About page for more information.

This article is based on research published under:

DOI-LINK: https://doi.org/10.48550/arXiv.2404.10111,

Title: From Predictive Algorithms To Automatic Generation Of Anomalies

Subject: econ.em

Authors: Sendhil Mullainathan, Ashesh Rambachan

Published: 15-04-2024

Everything You Need To Know

What is the significance of the "Allais paradox" in the context of economic theory and the use of machine learning?

The "Allais paradox" is a critical example that highlighted inconsistencies between expected utility theory and real-world decision-making. It demonstrated that individuals often make choices that contradict the predictions of expected utility theory when faced with lotteries involving risk. This paradox spurred the development of new economic theories. In this context, machine learning algorithms are employed to identify similar anomalies, accelerating the process of refining and improving economic theories by uncovering patterns that challenge established models. The algorithms help pinpoint areas where existing theories fail to accurately predict behavior, leading to a better understanding of decision-making under risk.

How do machine learning algorithms contribute to the identification of economic anomalies, and what are the implications of these discoveries?

Machine learning algorithms contribute to anomaly detection by analyzing patterns in data related to economic decisions, such as choices in lotteries. These algorithms create predictive models that can then be compared against existing economic theories. When the algorithm's predictions diverge significantly from what the theory suggests, an anomaly is identified. These discoveries challenge traditional economic theories and serve as a starting point for developing new theories that better explain observed behaviors. The implications are that researchers can refine existing theories and understand the nuances of decision-making, particularly in areas like asset pricing, game theory, and intertemporal choice.

Explain the process of how algorithms generate anomalies and how they are used to challenge existing economic theories.

The process involves using supervised machine learning to create a predictive model based on input features (x) and modeled outcomes (y*). The model's predictions are then contrasted with existing economic theories, which are modeled as mappings that return relationships. Anomalies are defined as the smallest set of examples where the theory's predictions are inconsistent with the algorithm's. The algorithm generates anomalies by framing anomaly searching as an adversarial game. A falsifier proposes features, the algorithm assesses these features, and the theory attempts to explain them. The anomalies generated are scenarios where the theory struggles to provide a meaningful explanation, thereby challenging the theory.

What are the key characteristics that define a robust economic theory, and how do machine learning algorithms help in evaluating these characteristics?

A robust economic theory should possess the following key characteristics: compatibility with data, consistency between data and theoretical predictions, refinement as more data becomes available, and non-trivial implications that provide meaningful predictions. Machine learning algorithms assist in evaluating these characteristics by identifying situations where a theory's predictions fail to align with observed data. The algorithm can identify inconsistencies, and the falsifier generates examples that challenge the theory to ensure it is tested under various scenarios. This process helps refine the theory and ensures it meets the criteria of robustness and provides more data to ensure better predictions.

How do these "black box" algorithms transform into tools that enhance economic understanding, and what is the ultimate goal of using them?

The supervised machine learning algorithms themselves are often described as 'black boxes' due to the complexity of their inner workings. However, the procedures in the text outline how these algorithms can be transformed into tools by automatically constructing anomalies from predictive models. By identifying the minimal instances where theory fails to align with algorithm predictions, researchers can scrutinize these anomalies. The ultimate goal is to accelerate the development of economic theory by revealing predictive signals and uncovering new empirical patterns that existing theories might miss. This helps economists refine existing theories and provides a deeper understanding of economic behavior and decision-making processes, especially under risky conditions. The aim is not just to revisit established concepts like risk and expected utility theory but to create a framework that can be applied to various economic problems.