Professor Vodenska Delves Deep on AI Ethics in Finance and Education
The only thing brought by the advent of generative artificial intelligence and large-language models greater than change is uncertainty. The very game-changing opportunity AI poses to all professional fields also poses potential drawbacks if the proper guardrails are not put into place. There are two realms where this is particularly true: finance and education.
As chair of the Department of Administrative Sciences at Boston University’s Metropolitan College (BU MET), Professor of Finance Irena Vodenska has given the vital dynamic between these two worlds more consideration than most. Having recently joined a Bloomberg podcast to discuss the uses of AI in international ESG policy, in the following interview, Dr. Vodenska offers insights into ethics in AI, from finance to education.
AI is pushing finance education to more closely resemble modern practice: data-driven, model-based, and explicitly ethical.
As she explains, AI use in areas like education and finance raises issues like bias, fairness, discriminatory patterns, systemic risk and their implications, skill erosion, and academic integrity versus authentic learning all demand careful consideration.
How can educators balance the needs of academic integrity and authentic learning with the powerful AI tools available at students’ fingertips? Who is responsible for the judgements rendered by AI? How can educators and financiers harness AI’s power while preserving essential values like human responsibility and fairness? Professor Vodenska answers these questions, while also providing additional insights into the unique convergence between AI, ethics, education, and finance. Read the interview below.
How do you define the role of artificial intelligence in higher education today?
For me, AI in higher education is both a mirror and a magnifier. It reflects the strengths and weaknesses of how we currently teach, and it amplifies whatever we choose to do with it. When a student can produce a passable essay or problem-set solution in seconds, it prompts we educators to ask: Were we genuinely assessing deep understanding, or were we primarily evaluating the ability to follow patterns and templates?
At the same time, AI magnifies our intentions. If we design rich, inquiry-based tasks, AI can help students explore more scenarios, see alternative perspectives, and get rapid feedback. If we rely on routine tasks, AI makes those tasks trivial and exposes their limitations. In that sense, AI doesn’t replace educators; it recenters us on the uniquely human parts of education: helping students develop judgment under uncertainty, ethical awareness, critical thinking, and a sense of meaning and purpose in their work.
At BU, we’re already living this shift. The AI@BU student modules and faculty workshops explicitly frame generative AI as something students must think with and think about: how it works, when it fails, where it lacks objectivity, and how to use it responsibly.
TerrierGPT, our BU-wide gateway to large language models, embodies this role: it provides students and faculty with secure access to multiple state-of-the-art models, and the pedagogy surrounding it emphasizes the critical use of these models, rather than blind trust.
How has the emergence and acceptance of AI begun to influence finance education?
Finance is almost an ideal testbed for AI. It is data-rich, model-heavy, and deeply concerned with prediction, risk, and systemic behavior. That reality is reshaping finance education in several ways.
- It is shifting the skill sets. It’s no longer enough for students to simply know the Capital Asset Pricing Model, the Black-Scholes option valuation model, or traditional Discounted Cash Flow techniques. They also need literacy in data pipelines, model evaluation, and the behavior of AI-driven systems. A valuation exercise today might ask students to compare a traditional DCF with an AI-generated scenario analysis, or to critique an AI-produced risk narrative using theory and data.
- Additionally, AI blurs the line between “quant” and “conceptual” finance. AI tools that can generate code, parse earnings calls, and summarize ESG or regulatory reports enable even non-programmers to work with richer and more complex datasets. The barrier to using models is lower; the bar for interrogating them is higher.
- AI helps us reframe model risk and systemic risk. We’ve usually worried about model risk and tail events in finance. AI adds opaque, adaptive models that can be embedded in trading, credit, or risk systems. Teaching finance now also means teaching how algorithmic behavior, portfolio management, and human bias interact in a socio-technical financial system.
In short, AI is pushing finance education to more closely resemble modern practice: data-driven, model-based, and explicitly ethical.
What parallels do you see between the rise of AI in financial markets and its rise in the classroom?
I see at least four strong parallels:
- Speed outpacing understanding.
Algorithmic and high-frequency trading grew faster than regulators—and sometimes traders—could fully understand. In the classroom, students adopted ChatGPT-style tools long before many courses or policies were updated. In both cases, practice outpaced theory and governance. - The Illusion of objectivity.
A model output, whether a risk metric or an AI-generated explanation, feels precise and authoritative, even when it rests on fragile assumptions. We now have to train both students and practitioners to ask: What data is this model trained on? What is the model ignoring? What incentives shaped the data and the model? - Concentration of power and resources.
In finance, the most advanced AI and data infrastructures are typically found in large firms. In education, advanced AI tools and computing resources are more readily available in well-funded universities and to students with greater financial resources. Without attention, AI can deepen structural inequalities in both arenas.
- Norms and regulation lagging behind innovation.
The metaphor I like to use here is, “Trying to catch a Ferrari by riding a bike,” because just as financial innovation typically precedes regulation, AI in education has progressed more rapidly than explicit norms regarding acceptable use. BU has established groundwork for incorporating AI use rules into syllabus statements, setting disclosure expectations, and developing a framework around AI, which is a work in progress. Working on these mitigation exercises is our attempt to close that lagging gap on the academic side.
These parallels are actually pedagogically useful. We can use what went right, and wrong, in algorithmic trading, for example, to help students think critically about AI in their own learning.
What ethical challenges does AI pose in both finance practice and finance education?
Many of the challenges are shared, even though they may surface differently in practice and in the classroom.
- Opacity and accountability.
In practice, when an AI-driven model denies a loan, moves capital, or triggers a trade, who is responsible for that decision—the model builder, the risk manager, the firm? Finance already struggles with “black box” models; AI intensifies this struggle. - Bias and fairness.
Financial data reflects historical inequities. If we train AI on that data without mitigation, we can encode and scale discriminatory patterns in credit scoring, hiring, or investment decisions. - Systemic risk and herd behavior.
If many institutions deploy similar AI models, they may react to signals in a correlated manner, thereby increasing the market’s fragility. BloombergGPT’s creators explicitly note that large, domain-specific models can shape many workflows; that power has systemic implications if misused or over-trusted.
In education, related ethical issues show up as:
- Academic integrity vs. authentic learning.
Students can now submit AI-generated work that appears polished but reflects a limited understanding of the subject matter. That’s why BU’s guidance emphasizes redesigning assessments, clarifying AI policies, and treating unacknowledged AI use as a form of academic misconduct, akin to plagiarism. - Equity of access.
Some students have access to powerful tools, high-end hardware, and fast networks; others do not. Without explicit support and the use of institution-provided tools like TerrierGPT, AI could widen achievement gaps. - Over-reliance and erosion of skills.
If students outsource thinking to AI, they may pass exams but be unable to reason under pressure. This lack of clear thinking under pressure is dangerous in finance, where their future decisions allocate capital and affect the livelihoods of people and families. The Ethics-in-Action and “over-reliance” modules exist precisely to address that risk.
So the core ethical challenge in both domains is the same: how to harness AI’s power while preserving human responsibility, fairness, and judgment. That is not a purely technical question; it’s about governance and values.
AI in finance is not just about models—it intersects with law, ethics, sociology, computer science, and sustainability. Co-designed modules or guest sessions can mirror the interdisciplinary reality of AI-driven financial systems.
What do you envision the “AI-powered classroom” of the next decade looking like?
What I don’t imagine is a classroom where an AI lectures while students passively watch. I envision AI as invisible infrastructure that supports a more human-centered learning experience.
- Before class, students interact with adaptive AI tutors that review prerequisite topics (time value of money, option pricing, and portfolio math) at their own pace, flagging misconceptions to the instructor. Tools like TerrierGPT, enriched with course materials, can provide BU-specific explanations and practice questions.
- During class we run richer simulations—say, a market with AI agents acting as traders, banks, regulators, and rating agencies. Students test strategies and immediately see system-level consequences, such as bubbles, liquidity crises, and contagion. Domain models like BloombergGPT (or its successors) ingest live news, filings, and macro commentaries in real-time, so the simulation is fed with realistic signals.
- Data and cases are dynamic. Instead of relying on stale textbook cases, AI helps us incorporate fresh data, news, and corporate disclosures into problem sets, allowing for more accurate and up-to-date analysis. Students ask natural-language questions about markets and then must cross-check AI answers against primary sources.
- Assessment is more authentic and continuous. AI systems provide ongoing formative feedback on reasoning quality—not just whether an answer is correct, but whether assumptions are explicit, evidence is adequate, and risk factors are identified. Instructors then focus more on coaching, mentoring, and facilitating discussion than on marking routine work.
Even in an AI-infused future, I think the heart of the classroom remains unchanged: a teacher framing the big questions, students challenging each other’s ideas, and a community collectively deciding what kind of financial system and what sort of AI-mediated society they want to build.
What advice would you give to finance educators just beginning to explore AI in their teaching or research?
A few practical suggestions:
- Start from learning goals, not from the tool.
Ask: What do I want students to be able to do that they currently struggle with? Then see where AI can help, richer scenarios, faster feedback, personalized practice, not the other way around. - Be open and experimental with your students.
You don’t have to brand yourself as “the AI expert.” It’s powerful to say: “We’re going to experiment with TerrierGPT (or similar tools) together and reflect on what works.” That models intellectual humility and lifelong learning. - Design the teaching to focus on the process, not just the product.
When you allow AI, ask students to disclose the intermediate steps, such as prompts, iterations, comparisons, and justifications. Ask them to demonstrate where they accepted or rejected AI suggestions and explain their reasoning. This framework makes reasoning visible and discourages one-click outsourcing. - Leverage institutional infrastructure.
Use tools like TerrierGPT that sit inside your university’s security, privacy, and policy framework rather than sending students to random external tools. This approach also allows you to integrate AI with other resources, such as course materials, BU modules on ethics and bias, and LMS-based activities. - Collaborate across disciplines.
AI in finance is not just about models—it intersects with law, ethics, sociology, computer science, and sustainability. Co-designed modules or guest sessions can mirror the interdisciplinary reality of AI-driven financial systems.
Above all: don’t wait for the “perfect” policy document. Start small, be transparent, and iterate.