The TraM project (Understanding and Supporting Traumatized Minor Refugees) is a BMBF project that is being implemented at ASH Berlin. Due to global conflict situations, more and more people are seeking refuge (European Commission, 2016), including many refugee minors. The goal of the TraM project is to support these minors and for this we need to understand their situation. At the same time, with vulnarable target groups, we also need to pay special attention to ELSI aspects (Ethical, Social and Legal Aspects).
Minors made potentially traumatic experiences in their home county or during the escape routes, that often takes month. Due to this, minors seeking refuge have a high risk of being traumatized and suffering from mental constitutions such as post-traumatic stress disorder (PTSD). To provide suitable help it is necessary to reach attended and unattended minors. To implement this, lower thresholds and a diagnostic model that fits the reality of life are needed.
It is a cultural issue weather seeking, getting and excepting help for mental support is accepted. The TraM Project aims to lower the threshold using a tree stage concept in order to bring young people to psychosocial support centers. The easiest way to reach young refugees is via mobile phone, because most of them have a mobile phone to reach their family and friends – independently where they are. The first step was to invent an artificial intelligence-based screening module. Here young refugees see pictures they describe for 20 seconds. During this time their movement (head and face) and the changes in their voice (voice pitch, intonation, arousal) are being recorded into data strings. Only these strings of data are the basis of the AI analyses. The result is reported to the refugee together with information about where to get help. One way to help is matching them with peer support that is included in the app and culture, language and age-appropriate material that is provided by an intelligent search and evaluation system. These steps are supposed to lower the threshold and make it easier for Minors to seek psychosocial support.
When dealing with these highly vulnerable group of people this project had to object Ethical, Legal and Social Impact more than other projects. First of all, the general idea of „Informed Consent“ is not enough when dealing with a highly vulnerable group of potentially traumatized minors. Additionally, you have to address the dual use dilemma when it comes to AI projects. On the other hand, it is verry important that we do research for vulnerable groups in order to provide the best possible support.
Since TraM project is an "integrated research project". This means that ELSI is not evaluated by external persons, but the scientists continuously evaluate the ELSI aspects and decide afterwards. One model that has proven to be very useful is the MEESTAR model (Manzeschke, 2015). The MEESTAR model is based on the principles of Value Sensitive Design (Friedman & Hendry, 2019) and from it 6 ELSI categories were derived.
The first categories is "privacy." These categories include items such as documenting all data access in writing and collecting data as sparingly as possible. It also includes keeping data on German servers and trying to keep as much data as possible in-house instead of outsourcing it to external parties. The second aspect is "Promote." This aspect includes diversity among the developers, training data that is similar to the target group and legal points (e.g. the Basic Law). The third aspect is "preventing harm." This includes regular functionality checks and updates. The fourth aspect is "autonomy." This point includes the traceability of the functionality and overridability. Overdriveability is a construct from autonomous driving and means that the human and his will are above the machine. The penultimate aspect is "fairness." This item includes evaluating for bias, checking for representativeness, and fairness. The last aspect is "transparency." This item includes the following aspects: Publication of the code, explanations of how it works, and choices for users.
Since the TRAM project is agile and changes are made constantly to meet the wishes and demands of the MR, ELSI aspects have to be considered constantly, too.
By repeatedly going through the process with external help (e.g., workshops on data ethics and critical whiteness), the principles of reflexive ethics are implemented. Reflexive ethics is very important in research and innovation (Kurtze & Wehrmann, 2016) in order to design products that are both appealing to the target audience and good for them.
The current status of the project and how we implement ELSI was presented at a conference (PSWB). Then the project was discussed with the scientific community. This discussion was another evaluation process. By repeatedly going through the process with external help (e.g., workshops on data ethics and critical whiteness), the principles of reflexive ethics are implemented. Reflexive ethics is very important in research and innovation (Kurtze & Wehrmann, 2016) in order to design products that are both appealing to the target audience and good for them.
In the discussion the following points were mentioned:
- Integrated research is complex
- constant reflection is important
- Vulnerable groups need a primary ethics vote and project-led ELSI evaluations
- continuous adjustments to project development is an extra effort that is important
- In projects with particularly vulnerable groups, the ELSI aspects are especially difficult but harder to meet
- Projects with vulnerable groups are important because we need in-depth knowledge to support them
Balmer, A. S., Calvert, J., Marris, C., Molyneux-Hodgson, S., Frow, E., Kearnes, M., & Martin, P. (2016). Five rules of thumb for post-ELSI interdisciplinary collaborations. Journal of Responsible Innovation, 3(1), 73-80.
European Commission. (2016). Horizon 2020 – The EU Framework Programme for Research and Innovation. Science with and for Society. Responsible Research and Innovation (RRI).
Friedman, B., & Hendry, D. G. (2019). Value sensitive design: Shaping technology with moral imagination. MIT Press.
Hasenbein, M. (2020). Ethik in Zeiten von Digitalisierung und künstlicher Intelligenz. In Der Mensch im Fokus der digitalen Arbeitswelt, 183-200.
Kurtze, H., & Wehrmann, C. (2016). Responsible Research and Innovation: reflexive Ethik in der Forschung - reflective ethics in research. Working Paper of the Institute for Innovation and Technology, 6.
Manzeschke, A. (2015). MEESTAR: ein Modell angewandter Ethik im Bereich assistiver Technologien. Technisierung des Alters–Beitrag zu einem guten Leben, 263-283.