← Back
Evidence source 6086Spot Checked

SAFE Artificial Intelligence in finance

Unknown venue2023-06-13Paper
Executive summary

Proposes Lorenz Zonoid-based metrics to assess AI trustworthiness in finance, focusing on sustainability, accuracy, fairness, and explainability.

What it examines

The paper proposes a set of integrated statistical methods using the Lorenz Zonoid tool to assess the trustworthiness of AI applications in finance, focusing on Sustainability, Accuracy, Fairness, and Explainability (SAFE). It aims to fill the gap in standardized metrics for evaluating high-risk AI applications.

What it concludes

The research offers a robust framework for evaluating AI applications' trustworthiness in finance, with potential applications in asset management, financial supervision, and market research. Future research could further refine these metrics and explore their applicability in other domains.

Extracted from this source

Evidence objects

Evidence 697875% extraction confidence
The research offers a robust framework for evaluating AI applications' trustworthiness in finance, with potential applications in asset management, financial supervision, and market research. Future research could further refine these metrics and explore their applicability in other domains.

key_findings bullet 1 · key_findings · validation V0

Raw abstract and provenance

Financial technologies, boosted by the availability of machine learning models, are expanding in all areas of finance: from payments (peer to peer lending) to asset management (robot advisors) to payments (blockchain coins). Machine learning models typically achieve a high accuracy at the expense of an insufficient explainability. Moreover, according to the proposed regulations, high-risk AI applications based on machine learning must be “trustworthy”, and comply with a set of mandatory requirements, such as Sustainability and Fairness. To date there are no standardised metrics that can ensure an overall assessment of the trustworthiness of AI applications in finance. To fill the gap, we propose a set of integrated statistical methods, based on the Lorenz Zonoid tool, that can be used to assess and monitor over time whether an AI application is trustworthy. Specifically, the methods will measure Sustainability (in terms of robustness with respect to anomalous data), Accuracy (in terms of predictive accuracy), Fairness (in terms of prediction bias across different population groups) and Explainability (in terms of human understanding and oversight). We apply our proposal to an easily downloadable dataset, that concerns financial prices, to make our proposal easily reproducible.

Source row: 1735 · abstract type: unknown