Interview with Andrea Martin

Short Bio

Andrea Martin is CTO Ecosystem & Associations for IBM in the DACH region. She previously served as CTO of IBM in Germany, Austria, and Switzerland and was President of the IBM Academy of Technology. She is a member of the German Science Council (Wissenschaftsrat), the Bavarian AI Advisory Council, and the Advisory Board of the TUM Center for Educational Technologies, with a particular focus on the responsible and trustworthy use of artificial intelligence.

Introduction Question

You began your career in applied mathematics and have since worked across technology leadership, innovation, industry, and public policy. Throughout this journey, what has remained constant in your understanding of how technology should serve people and society?

Andrea Martin: I have a huge passion for people AND technology, and the interaction between the two has been the constant throughout my career – in different ways of course. When I worked in consulting, it was the desire that people accept and productively work with technology that drove me. When I managed a global community of technical leaders (the IBM Academy of Technology) it was the enthusiasm of people to advance and progress technology for the value of everyone that I tried to motivate. And while being responsible for our IBM Innovation Studios in EMEA, I was always excited when we could inspire people with technology and make them imagine the next level of using technology for their employees and businesses.

Industry and Advisory Experience

Through your work at IBM, on the Advisory Board of the TUM Center for Educational Technologies, and in the Bavarian AI Advisory Council, where do industry, academia, and public policy most often misunderstand one another on AI and what would help them work together more effectively?

Andrea Martin: I don’t think they misunderstand each other. It’s rather different worlds, and everyone within these worlds has to make the effort of trying to understand and “being in the shoes” of those in
the other worlds. For me, transferring scientific or research results from academia into industry is one
of the major challenges – and it’s not a challenge because of misunderstandings, but possibly a lack
of understanding what the one can deliver and what the other really needs. Early collaboration between academia and industry could help overcome this challenge.

I don’t think they (industry, and academia) misunderstand each other […] but possibly lack understanding what the one can deliver and what the other really needs.
Andrea Martin

Similarly, I don’t think a lot of academic or industry leaders really understand how policy making works – and vice versa. For this reason, sometimes, regulations may seem impractical or industry may seem too intractable in complying to regulations. But if all parties worked together early in the process or at least were involved in reciprocal consultation and review processes, a lot of so-called misunderstandings could be resolved.

Drawing on your advisory experience, what would credible AI governance at a university look like in practice? Who should be involved, who should be accountable, and which decisions should always remain under meaningful human control?

Andrea Martin: As an industry representative, I’m not too familiar with the organizational structure of universities. However, governance should always be a joint task of several stakeholders. Those who provide AI solutions e.g., IT departments, those who use them e.g., educators or students, data privacy offices, legal departments, HR, etc. – all of them should be able to bring in their perspectives.
As an example: At IBM, we have a so-called Responsible Technology Board [1] with representatives
from all business units, which “oversees governance and decision-making around the development, deployment, and use of AI and other emerging technologies”.

For me, it’s another story which decisions should always remain under meaningful human control – this is rather a guardrail resulting from proper governance. In my view, AI should augment our intelligence and not replace it.
The EU AI Act gives us some guidance on what needs to remain under human control based on its risk classification e.g., anything that involves high risk for individuals. Other aspects are
goal setting, rules definition or deciding on ethical questions.

Insights concerning Education and Educational Technologies

Trust is a recurring theme in your work on AI. In education, what role should an EdTech Center play in building that trust? Should it prioritize evaluating technologies, developing governance standards, supporting educators and learners, or connecting these areas?

Andrea Martin: My short answer would be: The EdTech Center should do all of this. If the Center takes a comprehensive view – from concept to (recommendation for) selection of providers – if you don’t want to design and develop on your own – to governance and support for educators and learners, you can manage more or less the whole lifecycle of solutions and take appropriate responsibility.

What educational, technical, and ethical evidence should support an AI-based educational technology before it moves from a small pilot to institution-wide use?

Andrea Martin: The question is: Is there are difference in how AI applications move from pilot into production compared to any other applications? I’d say: not too much.
From a technical perspective you need to integrate the application into the existing architecture, considering security, access control, availability, performance, etc.
From a governance perspective you should monitor the expected value, consider ethical aspects including appropriate guardrails and human-in-the-loop and establish continuous monitoring of the application.
And of course you need organizational change activities such as training on the new application, managing resistance and turning it into acceptance, integrating the new application into existing processes, etc.

Looking Ahead

The TUM EdTech Center envisions personalized education through individually defined learning goals, continuous monitoring, and data-supported feedback. Given the reliance on algorithmic systems, what safeguards are needed to ensure that personalization empowers learners rather than becoming burdensome, controlling, or reinforcing existing inequalities through biased models?

Andrea Martin: In this context, transparency, robustness, data privacy, non-discrimination and accessibility are key aspects of responsible use of AI.
Transparency is important to let people know that the system they use is AI-based. Also, you should let people know based on which parameters they receive learning recommendations – so that they can possibly adapt those or provide additional information to get more suitable learning.
Robustness is important so that manipulation is made as unlikely as possible – both related to the documented learning results and the evaluation of learning results.
Also, personally sensitive data must be protected, as these personalized education systems can only work on personal data. But this data belongs to the user, and it should not belong to anyone else.
Anti-discrimination plays a role, as you want to treat learners in a fair way. For example, you don’t want the system to be biased and NOT to recommend technical learning experiences to women just because of their gender.
Last but not least, I want to mention accessibility. It’s important that all learners of a certain group gain the same access to personalized education in order to avoid (further) inequalities.

The TUM EdTech Center also seeks to make learning experiential through simulations, virtual reality and flexible digital environments. How can AI-supported immersive environments help learners safely explore decisions and consequences while preserving the social, embodied, and unpredictable dimensions of real-world learning?

Andrea Martin: Gamification can augment classical learning, as does experimentation: Hands-on experiences with technology or life science subjects can amplify theoretical learning. This does not mean that it’s the only way of learning. Therefore, I’m not afraid of losing the social or unpredictable real-world learning experiences – in my opinion, it’s a question of balance and also preference in learning styles.
The important thing is to not let the fear of unwanted consequences stop you in trying out new ways of learning. If it works, perfect! If not, you can stop using it.

The important thing is to not let the fear of unwanted consequences stop you in trying out new ways of learning.
Andrea Martin

As AI becomes more involved in feedback, assessment, and learning guidance, how can universities ensure that human relationships remain central to teaching and learning?

Andrea Martin: This topic has also been discussed in a work group of the German Science Council, with the findings, conclusions and recommendations now published at “Intellektuelle Souveränität: Empfehlungen für die Hochschulbildung in Zeiten von generativer KI” [2].
One of the recommendations is to establish AI-free zones – virtually or physically – in order to foster skills such as critical and transformative thinking, creativity and informed risk taking, which are maybe more important to some academic subjects than to others. These skills are important not only to preserve them as such, but also to be able to evaluate AI output.
By establishing these AI-free zones universities not only help building the skills mentioned above, but also enable social interaction, which is so important not only as an additional “skill”, but also relevant for self-development (including learning).

One of the recommendations is to establish AI-free zones […] in order to foster […] critical and transformative thinking, creativity and informed risk taking […] and self-development
Andrea Martin

You have mentored technical professionals throughout your career. What has mentoring taught you about how people learn and develop as well as how you lead and think about education?

Andrea Martin: Mentoring has taught me that some people learn because you tell them how to do things; that some people want to learn from your experience and see if it applies to them, too; some people learn from stories or things you point them to. Regarding the style, some people learn through interactions, others through face-to-face meetings, others through distant learning, videos, podcasts, etc. – there is a whole variety, and none of these learning ways and styles are better than the others. You just have to choose the ones right for you.
Mentoring has also taught me that it’s not a one-way street. Reverse mentoring i.e., when the mentors learn from their mentees, can be equally important than mentoring as such – especially as no one is an expert in everything. This is probably also a learning for educators or education in general: Don’t assume that the one standing in front of students knows more in all areas than the students. Be open, be curious, adopt life-long learning – that’s three aspects I have learnt throughout life and mentoring.

Andrea Martin
CTO Ecosystem & Associations for IBM in the DACH
Member of the German Science Council (Wissenschaftsrat)
Advisory Board Member, TUM Center for Educational Technologies

References
[1] https://www.ibm.com/think/author/ibm-responsible-technology-board.

[2] “Intellectual sovereignty: Recommendations for higher education in the era of generative AI”. https://www.wissenschaftsrat.de/SharedDocs/Pressemitteilungen/DE/PM_2026/PM_1526.