Humans in the Loop: Fixing AI Data and Creating Jobs for Displaced People
- yanabijoor
- May 14
- 3 min read
Updated: May 15
Founded in 2017 by Iva Gumnishka, Humans in the Loop tackles two major problems: fixing errors in AI data and creating steady jobs for people displaced by war or conflict.
The Problem: Bad AI Data and Lack of Work
AI tools rely on human-tagged data, especially for tasks like recognizing images. But this data often lacks standards, leading to mistakes and bias. The people labeling the data are usually paid very little and have poor working conditions, with few chances to grow their careers.
At the same time, more people than ever are being displaced by war and violence. In 2022, 103 million people were forced to leave their homes—a 15% increase from the year before, largely because of the war in Ukraine. Refugees face enormous challenges finding work, and typical aid programs only cover short-term needs.
Humans in the Loop connects these two issues by offering work to displaced people while helping fix problems in AI data.
The Solution: Quality Data and Fair Jobs
Humans in the Loop provides data labeling services for clients in health, farming, conservation, and city planning. Tasks include tagging images, drawing objects, and checking AI results. The team includes over 250 people, mostly women from countries like Syria, Ukraine, Iraq, and Afghanistan, who work from home, are paid fairly, and learn useful digital skills.
The company runs on a mixed model. A for-profit business provides the work, while a nonprofit offers training and job coaching. The company gives 25% of its profits to the nonprofit, creating a steady cycle of learning and earning. To keep things fair, Humans in the Loop avoids harmful labels (like race or gender stereotypes) and doesn’t work with clients connected to weapons or unsafe AI.
Iva started the project after volunteering at a refugee camp in Bulgaria and seeing how hard it was to find jobs. “The idea was to offer a simple job anyone could do so they could start earning right away,” she says. A partner in the AI space helped connect this need to the growing demand for people to check AI data, and a successful pilot led to the launch of Humans in the Loop.
Why It’s Different
Humans in the Loop stands out in three key ways:
Two-Part Impact: It improves AI data and provides real jobs beyond short-term aid.
Ethical Approach: It trains people carefully to avoid bias and follows clear rules to build safer AI.
Sustainable System: Its business and nonprofit support mix allows it to grow without relying on donations.
With staff in countries like Bulgaria, it delivers consistent, high-quality work, often better than larger crowdwork platforms. Unlike other short-lived projects, its self-funding model has helped it last longer than many older efforts.
What’s the Impact?
Jobs & Income: In 2023, the company hired 214 people, completed 162 projects, and paid out €421,795. Since 2017, it has given over $1 million to 952 people, mostly women, helping many achieve financial independence.
Training: 220 people have learned digital, coding, and design skills to prepare for other jobs.
Better AI: Projects include medical research (320 articles annotated for TrialHub), environmental work (5,500 images annotated for Restor), and health scans (thermal images annotated for Kelvin Health). The “Doctors in the Loop” project brings 29 doctors into AI work for better health results.
Global Reach: It now operates in places like Syria, Ukraine, and Kenya, and plans to grow to 1,500 workers by 2024.
What’s Next?
To grow its impact, Humans in the Loop is focused on:
Reaching More People: Better Internet and payment systems are needed in rural areas to help more workers.
Improving Tools: Adding real-time error-checking tools would help keep data quality high.
Expanding in Healthcare: The medical team could grow to meet the rising demand for medical AI annotation.
Tracking Progress: Following up with former workers could help show the program's long-term benefits.
Humans in the Loop is changing how AI data is crowdsourced while allowing displaced people to earn and learn. “There are so many people who need work,” says Iva. “We’ve been doing this for five years and we’re not planning to stop.” With more training, better tools, and the right partners, this project can grow even further, building a future where AI is fairer and more people have a way forward.
Sources:
Comments