A Mixed-Methods Study of Disputable Data in AI-Driven Financial Trading Environments
Lead Research Organisation:
London School of Economics and Political Science
Department Name: Methodology
Abstract
This project investigates the presence and impact of disputable data in AI-based automated trading systems. Inspired by earlier market disruptions such as the Flash Crash of 2010, the research seeks to understand how data irregularities that lack clear real-world correspondence may influence the behaviour of trading algorithms.
The study challenges the widely held assumption that the financial data used in automated systems is inherently reliable. It focuses on how disputable data is formed within the operational environments of financial data suppliers and why such data may go undetected, even in highly regulated markets.
The project addresses three core objectives: (1) identifying the types of disputable data present in historical quote feeds, (2) exploring the organisational and technical processes that contribute to the formation of such data, and (3) examining how existing quality assurance mechanisms may overlook these irregularities.
A mixed-methods approach is employed. Quantitative methods use statistical and machine-learning techniques to detect and categorise patterns within large-scale financial datasets. Qualitative methods, grounded in systems theory, draw on archival records to explore how organisational structures and systemic interactions shape data delivery processes.
The research aims to contribute to the field of Information Systems by advancing understanding of data governance, system reliability, and the risks associated with AI in critical financial infrastructures.
The study challenges the widely held assumption that the financial data used in automated systems is inherently reliable. It focuses on how disputable data is formed within the operational environments of financial data suppliers and why such data may go undetected, even in highly regulated markets.
The project addresses three core objectives: (1) identifying the types of disputable data present in historical quote feeds, (2) exploring the organisational and technical processes that contribute to the formation of such data, and (3) examining how existing quality assurance mechanisms may overlook these irregularities.
A mixed-methods approach is employed. Quantitative methods use statistical and machine-learning techniques to detect and categorise patterns within large-scale financial datasets. Qualitative methods, grounded in systems theory, draw on archival records to explore how organisational structures and systemic interactions shape data delivery processes.
The research aims to contribute to the field of Information Systems by advancing understanding of data governance, system reliability, and the risks associated with AI in critical financial infrastructures.
People |
ORCID iD |
| Maximillian Goehmann (Student) |
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| ES/P000622/1 | 30/09/2017 | 29/09/2028 | |||
| 2480084 | Studentship | ES/P000622/1 | 30/09/2020 | 29/09/2024 | Maximillian Goehmann |