Abstract
I have come to believe that a profit-driven society is not just
problematic, but incompatible with a better future. Much of the tech industry has become
increasingly pro-product and less pro-human, optimizing systems around growth and
performance rather than human needs. As a computer science student, I am acutely aware that
the systems my colleagues and I build shape how people learn, vote, spend, communicate, and
relate to one another. Yet these systems operate in a regulatory vacuum that allows firms to
define their own rules for data collection and algorithmic design, conditions under which Big
Tech thrives. My research begins from this reality: platforms monetize “free” services by
converting human behavior into data, while no comprehensive federal privacy law sets a baseline
for its collection, retention, or use. In parallel, engagement-optimized algorithms amplify
polarizing content, and AI systems have been trained on massive datasets, including copyrighted
material, with little transparency or accountability. Alongside this research, my technical
capstone project, hobski, explores what it looks like to build a digital platform centered on
people rather than extraction, connecting learners with experienced mentors in their local
communities.
Science, Technology, and Society (STS) frames engineering not as neutral
problem-solving but as world-building. An engineer needs to solve problems in the context of
the world in which they exist. This framing bridges the gap between technical applications and
society in a way that fosters a holistic perspective, promoting ethical responsibility. For example,
engineers can have all the skills to develop a complex interconnected ecosystem of chips to turn
humanity into an efficient hivemind, but STS allows us to question whether our individual
autonomous humanity is worth more. Using Actor-Network Theory (ANT) and Zuboff’s
political-economy lens, I analyze how code, data, policies, and corporations co-produce
outcomes. ANT exposes the networks of humans and non-humans, while political economy
explains why these networks tilt toward extraction and opacity under profit incentives. hobski
exists within this same platform ecosystem, yet it intentionally resists these dominant logics by
aiming to place transparency, consent, and human agency at the core of its design, treating trust
and privacy as foundational system requirements from the earliest research stages through
deployment.
The technical portion of my thesis focuses on the design, validation, and early
development of hobski as a peer-to-peer learning platform grounded in real human learning
behavior rather than abstract engagement metrics. To inform this work, I conducted hands-on
qualitative and quantitative research, including in-depth interviews and surveys, to understand
how people currently learn new skills and hobbies, what resources they rely on, and where those
resources fail them. This research revealed that while many people turn to scalable digital
resources such as YouTube or social media, they consistently report that learning from peers or
mentors is the most effective, yet also the hardest to access. These findings directly shaped and
are shaping the technical design of hobski, from its focus on local, human-centered connections
to its emphasis on reducing friction in discovering, scheduling, and participating in learning
sessions. Throughout the user research process, participants were clearly informed how their data
would be used, who would have access to it, and how their responses would inform the project,
establishing privacy and transparency. These same principles will govern data handling and user
interactions when the application is deployed.
The main problems I uncovered that result from the lack of regulations are the following:
(1) There is no federal baseline for cyber privacy, leaving most Americans governed by
corporate defaults. (2) Web-based tracking tools render “consent” largely fictional in practice. (3)
curation algorithms optimize for engagement, reliably boosting divisive content. (4) AI inherited
this playbook and is raising new intellectual property, transparency, and accountability risks.
In my STS research, I map how data extraction has become functionally unavoidable.
Drawing on audits of tracking infrastructures, I show that a few platforms achieve near-total
visibility into Americans’ digital lives; opting out of surveillance now often means opting out of
modern life. I hope to demonstrate how engagement-maximizing curation algorithms weaponize
those datasets in a self-reinforcing loop: each click fuels the model that serves the next, pushing
people into like-minded groups, even when no one is trying to be partisan. This explains why
small user interface tweaks cannot solve polarization if the incentive function remains
engagement, which translates to revenue.
In the paper, I express that opacity is an incentive outcome as well. Internal trade-offs
repeatedly privilege revenue over transparency, and the absence of comprehensive federal law
becomes an “actor” that locks in widely used systems that are expensive to reverse. I extend this
analysis to AI. I trace how permissive data norms of the 2010s seeded today’s IP and reference
crises and pushed the boundaries further.
Principles that should govern innovation:
1. Data minimization and meaningful consent by default: collection must be justified, not
assumed.
2. Source material and licensing obligations for commercial AI: traceable training data and
enforceable rights.
3. Align incentives with public values: move from a market that monetizes attention and
time to one that protects autonomy and guides innovation to be grounded in a desirable
future for all.
This is not merely a privacy or moderation problem; it is a governance problem. A
“free-market” guided by human values works only when those values are enforceable; without
that, asymmetric knowledge becomes asymmetric power over citizens and over democracy itself.
