Swansea University, April 2019
Human rights investigations are increasingly harnessing technology to assist the investigative process and to overcome some of the barriers faced in documenting and discovering human rights violations. Open source evidence is seen as having a hugely democratising potential, by allowing voices that have been historically been silenced in traditional investigations to be heard. However, it is possible that open source research creates informational biases of its own, in terms of the representativeness of the victims who have access to the requisite technology and the types of violations that are reported through open source evidence. There are also issues around the prioritisation of certain pieces of information by algorithms, and the story that those pieces of evidence tells and how this affects investigations.
These issues give rise to questions about the principles of participation, universality, equality and non-discrimination in the use of technology to advance and protect human rights.
This workshop examined the extent to which to which information gathered through open source research methods could complement and address some of the informational gaps inherent in traditional investigative methods, and whether the same or new informational biases arise in open source investigations. It examined whether harnessing technologically advanced methods for the collection and analysis of data can help to overcome any issues of informational bias that may arise in the context of open source investigations.
The specific objectives of the workshop were to:
- Examine how potential bias is identified in traditional investigations, and what steps are taken to mitigate it;
- Interrogate the methods through which informational bias in open source research can be identified and measured;
- Examine whether evidence gathered through open source research could complement and address some of the informational gaps inherent to traditional investigative methods;
- Ask whether open source research gives rise to informational biases of its own;
- Identify the extent to which automated tools for conducting open source research could overcome issues of informational bias that may arise;
- Share good practices, challenges and lessons learnt regarding the use of open source evidence for human rights fact-finding;
- Identify common trends and possible future directions for the conduct of open source research for human rights investigations, including means to make those investigations more systematic.
The workshop took place over two days, 10 and 11 April 2019. It was interactive, with short, informal presentations, structured in panel discussions in order to trigger discussions and debate. The discussions will look at relevant benchmarks, indicators and examples of best practice drawn from various perspectives. A ‘TechDemo’ session presented an early prototype of the Knowledge Hub Framework, an automated discovery tool that is being developed as part of the OSR4Rights project. The concluding session highlighted key conclusions and recommendations, and will discuss the way forward for discussions.
The workshop consisted of brief opening and closing sessions and the following panel discussions:
Panels 1 and 2: Representativeness in discovery and representativeness of violations
These sessions focused on the investigative process for human rights violations. They incorporated lessons learned from the perspective of NGOs as well as commissions of inquiry, fact-finding missions and other UN investigations on the representativeness of investigations. Combining these perspectives with lessons drawn from the academic literature on investigations and issues of representation, these panels set the scene for the remainder of the workshop.
Participants discussed whether issues surrounding the lack of representativeness in historic human rights investigations (e.g. lack of gender perspectives/focus on certain types of violations) are exacerbated or continued in investigations that rely extensively on open source research, or whether open source research may be able to overcome some of those issues.
Panel 3: Knowing what bias is and how to identify it
The objective of this session was to discuss what we mean by notions of ‘representativeness’ and ‘informational bias’ in the context of human rights investigations. As is well known from the literature, bias is often unconscious, but can have the adverse impact of silencing or overlooking particular perspectives. This panel examined types of bias that may arise in the conduct of human rights investigations – e.g. on the grounds of education, gender, race, age, digital literacy, urban v. rural divides, and/or algorithmic bias.
Panel 4: Overcoming issues of representativeness in open source research
This session looked to the future to examine whether and how the issues identified in the previous sessions can be overcome. It incorporated a technical demonstration of an early prototype of the ‘Knowledge Hub Framework’, a downloadable application to assist in the collection and processing of open source evidence. Participants examined the extent to which some of the automated methods incorporated through the Knowledge Hub Framework (and similar tools) could help to overcome some of the issues of representativeness that had been discussed.
The fruitful discussions from this workshop inspired a number of key outputs , including a JICJ article and an Opinio Juris blogpost, both co-authored by McDermott, Koenig and Murray. It also fed into two further journal articles, both currently under peer review.
This research was cited in relation to the ‘objectivity’ principle in the Berkeley Protocol on Digital Open Source Investigations, the international guidelines created by the United Nations Office of the High Commissioner on Human Rights and the Human Rights Center at UC Berkeley, published in 2020. McDermott contributed a module on this theme as part of the Amnesty International online course on Open Source Investigations for Human Rights.