The war in Ukraine brought more than 65,000 displaced people to Finland in 2022. Their matters were handled by authorities and support organisations with various information systems. How it really worked and how AI was involved has been investigated in a recent study by LUT University.
One clear finding is that refugee integration in Finland never runs on formal systems alone. Official platforms such as Enter Finland, Kela, and InfoFinland coexist with a dense layer of informal information systems: WhatsApp groups, Telegram channels, Facebook communities, word of mouth, and ad hoc ChatGPT queries.
“Both formal and informal information systems are important and should be more connected and consistently used. Informal channels supply speed, empathy, and native-language nuance that formal platforms structurally cannot, while formal platforms supply legitimacy and verification informal channels lack,” says LUT’s researcher Olena Ocheredko.
Rather than replacing informal networks with one official “super app”, Ocheredko would like future information systems developed in a way that connects these two layers, feeding verified information into trusted community channels and routing community-flagged misinformation back into official updates.
“AI and technology should not be used just to process refugees faster. It should help them – and the people supporting them – feel safe and understood, but most importantly find information,” says Ocheredko.
Official channels are rated positively despite everyday obstacles
The researchers conducted 35 in-depth interviews with Ukrainian adult refugees in 22 Finnish cities and 40 interviews with organisations such as Migri, Kela, and the Finnish Red Cross in early 2025. The interviews revealed both strengths and gaps in the management of services for Ukrainian refugees.
Almost all the refugees interviewed were highly educated, yet they ran into the same everyday obstacles, which were rooted less in a lack of ability than in a lack of accessible, coordinated information.
Identified strengths:
- Trust in official channels runs high. Most of the interviewed Ukrainian refugees rated especially Migri, Kela, and Info Finland positively. Once people find the right official information, they believe it.
- Humans are the glue. Volunteers, integration counsellors, church groups, and Ukrainian-led community organisers do the quiet, unglamorous work of translating, reassuring, and personal advocacy. They try to help in every possible way when Ukrainians are unable to use the digital systems on their own.
- Small, well-designed tools genuinely help. Video consultations through DigiVOK, hybrid language café events, and simple translation apps ease real friction and help build a human safety network.
Identified issues:
- Systems that wear people down. Opaque decision logic, identification problems, forms with no way to pause and come back later, unclear timelines for temporary protection status, and the need to repeat traumatic stories again and again all take a psychological toll.
- Translations are not understandable. Machine translation can render Finnish words into Ukrainian, but it can’t explain what those words mean institutionally. Many refugees still don’t understand what something as basic as municipality of residence actually determines for their lives.
- Trust must be earned, not assumed. Caseworkers and refugees alike remain cautious about AI touching anything related to legal status, personal data, or benefits, but they also mentioned that AI can help to find information.
AI is used by refugees as well as governmental and NGO workers
The research suggests that AI can play a pivotal role in areas such as language support, real-time information provision, and resettlement assistance. However, the ethical challenges surrounding transparency, bias, and the subjective nature of refugee claims must be carefully managed to ensure fairness and avoid exacerbating vulnerabilities. To address these issues, it is essential that AI systems are not only technologically advanced but also culturally responsive, trauma-informed, and inclusive in their design.
AI is constantly used in machine translation, searching for information, and refugees’ personal conversations. AI-assisted tools such as translators are also used by key personnel, including integration counsellors, health care and NGO staff, and volunteers. ChatGPT is used informally by caseworkers and refugees for quick answers, information searching, translation, and drafting.
However, AI is used unevenly: some caseworkers use it daily while others avoid it due to concerns about data protection and trust. Particularly the general data protection regulation (GDPR) was frequently cited as an obstacle to broader AI experimentation.
Recommendations for applying AI in global humanitarian contexts
Displaced people across the globe, such as asylum seekers, quota refugees, and people fleeing conflict or natural disaster, increasingly depend on AI and information systems, and so do the governmental organisations and NGOs that support them. Drawing on Finland’s evidence, Project Researcher Olena Ocheredko and Associate Professor Dominik Siemon give the following recommendations for deploying AI at scale in any humanitarian setting:
1. Keep AI in a supporting role, never a deciding one. Status determination, benefit eligibility, and protection decisions must stay in human hands. AI can translate, find, prepare, summarise, translate, and flag – never adjudicate.
2. Trauma-informed design for emotional and cognitive safety first, efficiency second. Interfaces should minimise cognitive load, allow pausing and resuming without penalty, and avoid repeated, high-stakes disclosure of trauma histories.
3. Build institutional and language clarity, not just raw machine translation. AI translation must be paired with plain-language explanations of legal terms, institutional roles, and system consequences. Literal translation without context increases fear of error when working with any information systems.
4. Formally recognise and connect informal networks. Refugees rely heavily on informal channels. Governments and agencies should treat these community channels as ecosystems to be managed and cross-verified, not bypassed in the name of GDPR compliance.
5. Institutionalise human mediation within digital workflows. Volunteers, caseworkers, and interpreters should be part of the formal system design, not a workaround for system failure.
6. Protect data with dignity, not just compliance. Clear, consent-based, multilingual explanations for what data is collected and why are essential to building trust among populations wary of surveillance.
7. Co-design with the people the system serves. Refugees and frontline workers should help shape information system tools from the start – participatory trauma-informed design consistently outperforms top-down deployment in trust and uptake.
8. Inclusive access must be prioritised through multilingual interfaces and digital literacy support. Systems should embed transparent decision-making and clear appeal mechanisms for users affected by AI outputs. Participation must be institutionalised by creating compensated advisory roles for refugees and with refugees and by establishing iterative feedback loops during the development and rollout of new digital systems.
LUT’s research project “AI and Information Systems in Humanitarian Contexts: A Theoretical and Empirical Analysis of Refugee Management in Finland” will be completed by January 2027. The project is funded by the Research Council of Finland.
Visit the project website for further information and published papers.
More information:
Dominik Siemon
Read next:
The Curious People newsletter shares our solutions for helping build resilient communities, industry, and businesses while promoting the energy transition and the regenerative use of natural resources.