Ethics | Digital Finance MSCA

Work Package 9
Ethics in DIGITAL

Leading institution

Useful links

WP9 Github

Overview

Work Package 9 is dedicated to pioneering ethical practices in financial data analysis through the application of Explainable AI (XAI). In an era where data privacy and ethical AI are paramount, our initiative stands at the forefront of integrating ethical reflection into cutting-edge technological advancements. This page offers insights on the ethical standards and objectives within the MSCA DIGITAL project.

Why an Ethics Mentor is Essential

In this MSCA DIGITAL project, doctoral students will delve into the analysis of financial transaction data, which inherently contains detailed and sensitive information. Given the specificity and temporal scope of this data, anonymization is impractical, rendering the data pseudonymous and, therefore, subject to stringent GDPR regulations. Our commitment is to ensure that our research not only adheres to legal standards but also embodies the highest ethical principles.

WP 9 Team

Catarina Silva

WP9 Lead

Purpose

Promote Ethical AI: Train participants on the ethical dimensions of financial data analysis, focusing on privacy, fairness, and trust.
Integrate Ethical Reflection in XAI: Ensure that the development and application of XAI systems are ethically sound, providing transparent and fair decision-making processes.
Enhance Awareness: Instill a deep understanding of the ethical challenges and responsibilities involved in handling financial data and AI technologies.
Identifying issues during reviews

WP9 is also responsible for identifying ethics categories and issues during reviews on the topics shown below

Doctoral candidates responsibilities

DCs are responsible for anonymizing restricted data before sharing it with their supervisory team. This will adhere to the requirements of the research ethics committee and legal counsel service of the beneficiary and permit online access to databases via NDA and VPN tools.

Ethical dimension of the objectives, methodology and likely impact

The present project proposal deals with the issue of Digital Finance. This means that it combines access to data and data issues with AI aspects. Consequently, the main ethical dimensions that are found in this proposal are related to personal data and the use of AI. For a successful implementation of the action in order to address the project's objectives, data needs to be manipulated and analysed. There is the need for data from financial actors like banks, regulators, or other financial companies. All this data will be processed using two main types of methods: statistical/ econometrics and AI. Ethical issues related to personal data that can be attached to our objectives are given by the collection of data, organization and storage, adaptation/ alteration, retrieval and consultation, use, and transferability/ sharing. We will use both available data related to individuals and financial actors, and data that will be collected through surveys or interviews. All members of the projects will consult databases with personal data. For survey reasons, mailing lists will have to be constructed. Additionally, in order to increase the impact of our results we will ask partners and respondents if they would like to provide their e-mail addresses or other types of personal contacts to be kept up-todate about the results of our research activities. The topic addressed by the present project proposal does not lead to any harm in the health or the physical safety of the participants. It may, however, hinder their safety, from the perspective of their personal data being employed in the research.

The second major ethical dimension that we emphasize for our project is the use of AI. Digital Finance is characterized nowadays by a substantial amount of data (Big Data). Econometrics tools have to be combined with the use of AI methods, in order to obtain the most accurate results. AI techniques will be both used and developed to ensure the objectives' fulfillment. But, just as expressed by the European Commission, there are different ethical aspects that have to be taken into account. One very important is related to the trustworthiness and robustness of AI. AI models/systems can generate confusion or affect human autonomy. Additionally, both human and other artificial actors can interact in an adversarial manner with the developed AI system. Our goal is to fill in the research gap existing in the field of Digital Finance and Explainable AI. For this, we will also provide strong interdisciplinary training for future European financial and data researchers in Europe, in order to boost capabilities and create a qualified workforce. Consequently, ethical dimensions are also to be found in the training programmes.

Another important aspect is that some of the participants in the project proposal come from non-EU countries - Switzerland. Consequently, additional ethical issues may arise, depending on the national legislation of the beneficiary. This will all be treated according to the EU ethical guidelines.

Compliance with ethical principles and relevant legislations

To address the ethical issues mentioned in the Ethics Self-Assessment part in the Grant Agreement, we will first adhere to the highest ethical standards provided by both national and European regulations in the field. In each particular case (non-EU countries participants included) the strictest ethical regulation will be applied and all participants will have to implement it.

In respect to the AI ethical issues that was emphasized in the Ethics Self-Assessment, the first step is to use the ALTAI checklist, as stated before. To counteract the adversarial usage of AI, all applications developed within the project will be safely kept and used according to international laws. Additionally, we will take care that AI tools that we use or develop preserve the human decision-making autonomy and that they are only a means to achieve better results such as to contribute to the fulfillment of the project's objectives, on one hand, and to an increase in the standard of living of individuals, by increasing financial innovation and financial access. We will involve an advisory board made up of specialists in the field. To mitigate this risk, all tools and applications will be tested several times and in comparison to other types of approaches, like, for example, Econometrics. All activities related to the use of AI will comply with the ALTAI checklist and the 7 major requirements in this list.

Network Organization.

The consortium is confirming that compliance with ethical principles and applicable international, EU and national law in the implementation of research activities not originally envisaged (or not described in detail) in the DoA will be ensured. Any ethical concerns raised by those activities will be handled following rigorously the recommendations provided in the European Commission Ethics Self-Assessment Guidelines.

The consortium is confirming that in case of an Ethics Check, relevant documents, authorisations, approvals will be obtained, kept on file and sent to the REA upon request.

D9.1. Appointment of Ethics Advisor (Due in M6)

This project should include an Ethics Mentor for the following reason:

The proposal intends to analyze financial transaction data of persons by doctoral students. Because of the typical detail of such data and the time frame encompassed it is not possible to effectively anonymize this data. Hence it needs to be assumed that the data is pseudonymous data which is considered as personal data according to the GDPR. In this context explainable AI (XAI) is going to be investigated which allows insights in the decisions of the classifier and needs to be ethically reflected. The proposal states that the aim is "explainable and fair AI-generated decision". The consortium shows some awareness by mentioning a potential member of the advisory board with expertise in this area but in the context of the training measurements these subjects are only briefly sketched. Also in the expertise of the scientific advisors these subjects are not substantially present.

The Ethics Mentor should train participants on the ethical dimensions including privacy, fairness and trust in the context of financial data and the intended financial data space. This training should be given to participants at the latest in month 12. In the training concerning XAI (M12) in addition an ethical reflection needs to be introduced. Also in the training course concerning sustainable finance (M18) a critical reflection of the application of XAI and blockchain should be introduced.

For the doctoral students the Ethics Mentor should provide reflection of the ethical dimensions of the individual research projects.

These measurements needs to be documented in a report at month 24 and another report at the end of the project.