Ecorys conducts feasibility study on a forecasting and early warning AI tool for migrationClient: European Commission - DG Migration and Home Affairs (HOME) | Sectors: Security and Justice
During the so-called migration crisis in 2015-2016, the European Union (EU) and its Member States found themselves insufficiently prepared for the arrival of large numbers of migrants at their borders. This prompted the EU and the Member States to invest in improving their capacity to better anticipate and respond to future migration flows. As part of this, DG Migration and Home (HOME) requested Ecorys to conduct a feasibility study on a forecasting and early warning tool for migration based on Artificial Intelligence (AI) technology.
Identification of requirements for AI-tool
Practically, the AI-Tool would incorporate and adequately process various data sources on all stages of irregular migration. This includes assessments of situations in third countries in the first place, which could provide early indications of onsetting movements of irregular migration. To investigate to what extent it is feasible to develop such a tool, the study team undertook various research activities.
Firstly, the researchers conducted desk and literature reviews to conduct initial scoping interviews with key stakeholders from various EU agencies and bodies. The researchers followed up with expert interviews to be able to carry out a general and operational assessment of the AI tool. A legal assessment, trustworthiness assessment and organisational assessment followed. Taking the assessments into consideration, the study team developed a case study to contextualise and exemplify the study’s findings further and concluded the study with a roadmap to action, including recommendations for DG HOME.
Feasibility of the AI-tool
The study team found that there is sufficient data available to inform the AI tool and that no significant legal obstacles exist regarding the data needed to inform the tool. In addition, they found that the appropriate AI architecture can be developed. However, obstacles arise regarding the development of the tool in general, as amendments to secondary EU legislation might be necessary. Also, the hosting structure of the AI tool is feasible but would require investments into additional resources. However, the most prominent obstacle is the lack of a governance structure to facilitate appropriate coordination mechanisms for the AI tool, such as additional working agreements and a central coordination point to facilitate data access and data sharing.