Alexios Mantzarlis profile interview
Alexios Mantzarlis is the Director of the Security, Trust, and Safety Initiative at Cornell Tech. We recently asked Alexios a few questions for our new Responsible Tech Guide.
“If I have any advice to give, it is to those interested in T&S without an engineering background. You can bring to Trust & Safety whatever expertise you have developed; it is valuable. But you will have to find a way to gain practical experience of the technology itself.”
Describe your current role. How does your work address complex technology and society challenges?
The Security, Trust, and Safety Initiative (SETS) is a hub for cybersecurity, privacy, and Trust & Safety at Cornell Tech. We help prevent and mitigate digital harms through graduate education, active communities of practice, and research.
My primary goal with SETS is to help Cornell Tech students develop a well-rounded understanding of how crucial digital safety is to well-functioning societies -- and how reckless product design and platform management can endanger them.
I pursue this goal primarily through (i) a technical Trust & Safety course and certificate heavy on ethical frameworks and hands-on activities (ii) a Red Team clinic that supports nonprofits deploying AI for good, and (iii) a broad effort to engage industry leaders with students through talks, panels, and research projects.
My aspiration is that SETS graduates are both able and motivated to place digital safety at the heart of whatever career they end up pursuing in tech -- and that this will make our online spaces a bit better for everyone.
Tell us about your career journey. What key steps brought you to your current role, and what guidance would you give someone following a similar path?
I worked in government, think tanks, consulting, and journalism before falling into Trust & Safety.
My "a-ha" moment came in 2016 when I was leading the International Fact-Checking Network (IFCN). The IFCN was set up to promote this form of accountability journalism aimed at increasing the reputational cost of public lying and enabling citizens to get the facts in a noisy public discourse.
It quickly became evident that platform dynamics meant some of those accountability mechanisms were broken by algorithms optimizing for engagement rather than information quality. Working with Facebook to launch the Third Party Fact-Checking Program (RIP) was a great lesson in how much platforms can do to mitigate the proliferation of baseless hoaxes. The subsequent 8 years -- which I spent at the IFCN, Google, and Cornell Tech -- served as a reminder that information infrastructure is a socio-political challenge pretending to be a technical one.
If I have any advice to give, it is to those interested in T&S without an engineering background. You can bring to Trust & Safety whatever expertise you have developed; it is valuable. But you will have to find a way to gain practical experience of the technology itself. One great thing about the AI Red Teaming wave is that anyone can do it without permission or fancy tools. I encourage students to (ethically) build a portfolio of safety tests on tools they are familiar with and share it with recruiters.
What are the biggest barriers you’ve encountered in advancing responsible technology within your field, and how have you addressed them?
Business goals can sometimes get in the way of responsible technology. Political motivations can often do that too. But I think the biggest challenge in advancing T&S goals I encountered was the lack of sound metrics to counter opposing arguments.
It is often hard to precisely quantify the impact of a great Trust & Safety intervention at a user, platform, or society level. Some of it comes down to measuring a counterfactual -- how do you know exactly how much damage didn't happen thanks to a content policy? -- but some of this is also because the right data isn't being collected.
It's a little trite to say that if you can't count something, it risks not getting counted, but that has been my experience.