Nico Ajzenman
@nicolasajz.bsky.social
Econ Assistant Professor @McGillU. Affiliated to
@IZA_bonn and @JPAL 🇦🇷 in 🇨🇦.
Newsletter - Esto no es Economía: http://www.noeseconomia.com
www.ajzenman.com
@IZA_bonn and @JPAL 🇦🇷 in 🇨🇦.
Newsletter - Esto no es Economía: http://www.noeseconomia.com
www.ajzenman.com
And this is the link to our first paper (with great Bluesky-less coauthors, of course) www.journals.uchicago.edu/doi/10.1086/...
Humans versus Chatbots: Scaling-Up Behavioral Interventions to Reduce Teacher Shortages | Journal of Political Economy Microeconomics: Vol 0, No ja
www.journals.uchicago.edu
July 9, 2025 at 2:03 PM
And this is the link to our first paper (with great Bluesky-less coauthors, of course) www.journals.uchicago.edu/doi/10.1086/...
About tone: We're now analyzing empathy—are AI bots more or less empathetic than humans? And how persuasive is Kai?
Coming soon: Did Kai perform as well as human counselors in increasing applications to teaching programs?
More about what we've learned here blogs.iadb.org/educacion/en...
Coming soon: Did Kai perform as well as human counselors in increasing applications to teaching programs?
More about what we've learned here blogs.iadb.org/educacion/en...
From human counselors to AI agents: What We Learned From Scaling Career Guidance in Chile - Enfoque Educación
Scaling effective educational interventions remains a major policy challenge. Programs that succeed at a small scale often lose impact when expanded—a phenomenon researchers call the "voltage effect."...
blogs.iadb.org
July 9, 2025 at 2:03 PM
About tone: We're now analyzing empathy—are AI bots more or less empathetic than humans? And how persuasive is Kai?
Coming soon: Did Kai perform as well as human counselors in increasing applications to teaching programs?
More about what we've learned here blogs.iadb.org/educacion/en...
Coming soon: Did Kai perform as well as human counselors in increasing applications to teaching programs?
More about what we've learned here blogs.iadb.org/educacion/en...
About content:
- Most Kai’s interactions were factual queries (course info, scholarships), while human counselors handled more subjective questions (e.g. "Am I patient enough to teach?").
Students did not interact the same way with humans or bots
- Most Kai’s interactions were factual queries (course info, scholarships), while human counselors handled more subjective questions (e.g. "Am I patient enough to teach?").
Students did not interact the same way with humans or bots
July 9, 2025 at 2:03 PM
About content:
- Most Kai’s interactions were factual queries (course info, scholarships), while human counselors handled more subjective questions (e.g. "Am I patient enough to teach?").
Students did not interact the same way with humans or bots
- Most Kai’s interactions were factual queries (course info, scholarships), while human counselors handled more subjective questions (e.g. "Am I patient enough to teach?").
Students did not interact the same way with humans or bots
Some early learnings analyzing conversations:
- Kai led with 35% initial engagement > than human counselors (28%)
- Kai responded faster & more factually focused. Kai and humans covered = number of topics per conversation.
- Kai seems more efficient: simillar info, similar engagement, shorter time
- Kai led with 35% initial engagement > than human counselors (28%)
- Kai responded faster & more factually focused. Kai and humans covered = number of topics per conversation.
- Kai seems more efficient: simillar info, similar engagement, shorter time
July 9, 2025 at 2:03 PM
Some early learnings analyzing conversations:
- Kai led with 35% initial engagement > than human counselors (28%)
- Kai responded faster & more factually focused. Kai and humans covered = number of topics per conversation.
- Kai seems more efficient: simillar info, similar engagement, shorter time
- Kai led with 35% initial engagement > than human counselors (28%)
- Kai responded faster & more factually focused. Kai and humans covered = number of topics per conversation.
- Kai seems more efficient: simillar info, similar engagement, shorter time
FF: now there are AI-powered bots. We created Kai—a 2024 generative AI chatbot on WhatsApp designed for more tailored, adaptive interaction. We trained them for many months. In a new RCT with 43,000+ students, we compared: Human counselor versus Kai (AI agent) vs control.
July 9, 2025 at 2:03 PM
FF: now there are AI-powered bots. We created Kai—a 2024 generative AI chatbot on WhatsApp designed for more tailored, adaptive interaction. We trained them for many months. In a new RCT with 43,000+ students, we compared: Human counselor versus Kai (AI agent) vs control.
It's very human intensive. In the same experiment (2020) we used the technology available back then to try to scale it up: scripted bots. They achieved only ⅓ to ½ of those results and they were insignificant.
They lack the engagement and personalization essential in counseling.
They lack the engagement and personalization essential in counseling.
July 9, 2025 at 2:03 PM
It's very human intensive. In the same experiment (2020) we used the technology available back then to try to scale it up: scripted bots. They achieved only ⅓ to ½ of those results and they were insignificant.
They lack the engagement and personalization essential in counseling.
They lack the engagement and personalization essential in counseling.
Elige Educar’s “Quiero Ser Profe” program in Chile tackles teacher shortages by offering personalized WhatsApp counseling to highschoolers, encouraging them to pursue teaching.
It works (JPE Micro forthcoming). It increased actual enrollement in teaching majors.
BUT...
It works (JPE Micro forthcoming). It increased actual enrollement in teaching majors.
BUT...
July 9, 2025 at 2:03 PM
Elige Educar’s “Quiero Ser Profe” program in Chile tackles teacher shortages by offering personalized WhatsApp counseling to highschoolers, encouraging them to pursue teaching.
It works (JPE Micro forthcoming). It increased actual enrollement in teaching majors.
BUT...
It works (JPE Micro forthcoming). It increased actual enrollement in teaching majors.
BUT...
With R Perez Truglia, G Cruces, D Tortarolo and G Vazquez Bare
February 20, 2025 at 1:58 PM
With R Perez Truglia, G Cruces, D Tortarolo and G Vazquez Bare