Introduction (Babak)
Hello everyone. In this presentation, we’ll explore the economic impact of Duolingo’s decision to
become an “AI-first” company. Using AI, Duolingo accelerated the production of its language
courses, creating 148 new ones in under a year, compared to the 12 years it took to make the
first 100. While this shift saved money and boosted output, it also resulted in job losses, raised
questions about quality, and changed the platform’s value in the education market. We were
given an AI-generated economic analysis of this shift and were asked to evaluate it using the
economic concepts we’ve learned in class. Today, we will be comparing our analysis to the AI’s
version and then introducing our solution to help Duolingo remain competitive and socially
responsible.
SLIDE
Needs, Wants, Scarcity, and Resources (Charlie)
The AI analysis claimed Duolingo is just a “want” — something nice to have but not essential.
We expanded this by pointing out that for migrants trying to learn a language in a new country,
for international students, and for workers needing language skills for employment, Duolingo
can be a need. It’s not just about fun — it’s about economic mobility and integration. The AI
correctly noted that human labor in course development is scarce, especially since it used to
take years to create high-quality courses. However, it failed to recognize the deeper scarcity:
culturally accurate and educationally sound content. Language learning is more than translation
— it’s about nuance and context, and AI doesn’t always get that right. Lastly, the AI called user
data a scarce economic resource, which is incorrect. Data is not finite or rivalrous in the same
way land, labor, and capital are. It’s better classified as an input or byproduct, not a traditional
economic resource.
SLIDE
Incentives and Unintended Consequences (Babak)
The AI identified incentives like cost savings, faster production, and investor interest. These are
true and reflect standard corporate goals to maximize output and minimize cost. But it left out
some unintended consequences. One of the biggest is cultural dilution. AI-generated content
can easily miss idioms, local expressions, or social norms, especially for minority or Indigenous
languages. Another consequence is loss of worker morale and structural unemployment. These
were highly skilled workers who were translators, educators and voice artists who were not
easily replaced. Their job loss isn’t just financial; it’s emotional and professional. Over time, this
can reduce innovation, discourage students from entering the field, and hurt the broader
economy.
SLIDE
Supply, Demand, and Equilibrium (Charlie)
Duolingo’s use of AI significantly shifted the supply curve. It created far more content in a
fraction of the time and at a lower marginal cost. In the short term, the demand for the app likely
increased as users were attracted to the variety of new courses and languages. So we see both
supply and demand moving right, and equilibrium quantity increasing.
But in the long term, problems arise. If users find that AI-generated lessons are repetitive,
inaccurate, or culturally off, they’ll leave. Worse, many may skip Duolingo entirely and use tools
like ChatGPT directly, which are free and customizable. This could shift the demand curve left,
creating a market surplus, where there is more content than users actually want. Even though
the app is free, this leads to wasted resources, lowered user engagement, and reduced brand
value. We modeled both of these outcomes in two supply and demand graphs: one showing
short-term growth and another showing long-term decline and surplus.
SLIDE
Globalization and Technology (Babak)
Duolingo’s use of AI has dramatically improved its global accessibility. Previously, users needed
English as their base language. Now, thanks to AI, someone who speaks Hindi, for example,
can learn a language like Korean directly. This is a huge step in making language education
more inclusive and global. It boosts human capital, particularly in areas that were previously
underserved by traditional language education.
However, the AI analysis missed a critical cultural risk. Most AI systems are trained on data that
is heavily biased toward English and Western norms. This creates a danger when developing
courses for non-Western languages. Poor translations, missing cultural context, or even
offensive errors can occur. These risks not only hurt users but also damage trust in the platform,
especially in communities that already face digital exclusion.
SLIDE
GDP and Economic Indicators (Charlie)
The AI analysis correctly pointed out that automation increases productivity. Duolingo is now
producing more with fewer inputs, which boosts output and sectoral GDP in the short term. It
also improves long-term human capital by making education more accessible.
However, what the AI didn’t discuss is how this affects aggregate demand. When workers are
laid off, household income goes down. That means less spending and less consumption, which
can drag GDP in the short run. The benefits of automation often go to capital owners —
investors and executives — while workers face financial strain. This deepens inequality and
reduces the overall GDP multiplier effect. So while Duolingo’s AI shift looks good on paper, it
introduces demand-side risks that can’t be ignored.
SLIDE
Rejected Solution (Babak)
When we first started thinking about solutions, our first idea was to try to reverse some of the
effects of automation by bringing human workers back into the content creation process. We
thought that maybe Duolingo could create a separate premium track of language courses made
entirely by professionals — real translators, educators, and linguists. It felt like a respectful way
to protect jobs and give users a more “authentic” experience. But as we thought more deeply
about the company's goals and the direction of the edtech industry, we realized this approach
would create more problems than it solves. It would be expensive, inefficient to maintain
alongside automated production, and unlikely to attract enough users to justify the high costs.
SLIDE
Our Two-Part Solution (Charlie)
Now let’s talk about our proposed solution. We believe Duolingo can maintain its AI-first model
while also addressing both labor displacement and long-term competitiveness — through a
two-part solution.
SLIDE
Part One: Worker Redeployment to AI-Focused Roles
Instead of simply laying off workers, Duolingo can reassign them to new AI-supporting jobs.
These include:
● AI Content Reviewers, who check for accuracy and cultural sensitivity.
● Prompt Designers & Testers, who refine the instructions AI uses to generate content.
● Voice Checkers, who make sure AI pronunciation sounds natural and native-like.
This helps Duolingo avoid costly mistakes and keeps trust in the platform high. It also reuses
the expertise of skilled language professionals rather than wasting it. We fund this by redirecting
a portion of the money saved from not hiring contract workers into retraining and onboarding for
these new roles.
SLIDE
Part Two: Certified Learning Pathways (Babak)
The second part of our solution helps Duolingo stay relevant in a future where people may turn
to free AI tools like ChatGPT and Deep Seek instead. Our idea is to partner with universities
and language organizations to offer certified courses. When users complete a course, they
receive an official certificate that could be recognized by schools, employers, or immigration
programs.
Unlike free AI tools, Duolingo would then offer something unique — academic and professional
legitimacy. This gives users a reason to pay for Super Duolingo and continue using the app
long-term, especially if they want to prove their skills on a resume or in a visa application.
SLIDE
Strengths
● Keeps experts employed and boosts trust with verified content.
● Certification adds real value over free AI tools.
Weaknesses
● Requires upfront investment and time to scale.
● Not all workers may adapt to new roles.
Opportunities
● Leads in human-AI hybrid learning.
● Attracts students, professionals, and institutional partners.
Threats
● Rivals may copy the model quickly.
● Trust or oversight may decline if poorly managed.
SLIDE
Conclusion (Babak)
To wrap up, Duolingo’s shift to AI gave it speed and efficiency — but at a cost. Jobs were lost,
course quality was questioned, and demand could fall in the future. We believe our two-part
solution — combining worker redeployment with official certification — offers a sustainable,
ethical, and strategic way forward. It supports workers, builds user trust, and gives Duolingo a
long-term edge in a competitive market. Thank you.