Symposium on Visual Investigations: Best Practice in Collaboration and Community-led Approaches – Data Design and Analysis for Human Rights Audiovisual Archives – Airwars and Bangladesh Protest Archive

[Clive Vella is Data Manager at Airwars, a not-for-profit transparency watchdog which tracks, assesses, archives and investigates civilian harm claims in conflict-affected nations. 

Shoeb Abdullah is a digital rights activist and OSINT investigator based in Bangladesh. He is co-founder of the Bangladesh Protest Archive, an initiative dedicated to preserving, documenting and investigating gross human rights abuses around democratic protest. 

Kartika Pratiwi is a communications and development specialist with experience using visual storytelling, documentation, and collective memory for social change. At WITNESS, she works across the Asia-Pacific region, particularly in Bangladesh, West Papua, and Indonesia, supporting frontline defenders and communities to use video and technology for human rights documentation, advocacy, and accountability.]

In early 2026, WITNESS, in collaboration with Airwars, initiated five workshop sessions, led by Clive Vella (Data Manager, Airwars) to train a core team within the Bangladesh Protest Archive (BPA) in the designing of their datasets. This training was delivered using Airwars’ methodology, developed through its extensive experience documenting civilian harm and structuring open source evidence for investigations, while building on the BPA’s collection of textual and audiovisual documentation from the 2024 Bangladesh protests. This effort comes as part of a wider project by Airwars that has been recently funded by the Swedish Postcode Lottery Foundation which enables the organisation to enhance access to reliable and comparable information on the civilian impact of war. These workshops were facilitated by Georgia Edwards with ad hoc support from other WITNESS colleagues. 

The Bangladesh Protest Archive is a community-led initiative established by Bangladeshi activists, with support from WITNESS and in collaboration with Activate Rights, to preserve, verify, and make accessible documentation of human rights violations committed during the 2024 Bangladesh protests. WITNESS partnered with Airwars because of its extensive expertise in developing structured methodologies for organizing open source evidence and supporting investigations, helping BPA transition from collecting documentation to building a sustainable, investigation-ready archive for accountability.

Airwars’ core conceptual approach was informed by their own experience: designing a dataset with clear definitions and with specific goals in mind, while also understanding who are the people that need to see and understand your data is essential to effective advocacy. Through experience, we have understood the value of applying data management concepts at the very beginning of a data-dominant project as it is inevitable that in the later stages, when analysis and sharing becomes paramount, the quality of data, the ability to clearly define it, and the ability to share and proliferate are critical to the success of the organisation.

It is important to emphasise the terms “share” and “proliferate” because these are challenges that go beyond the act of simply putting your work online. As organisations, especially small ones, operate daily under the tension of their mission, or in many cases, under the threat of violence and death, streamlining data processes can appear as an additional task rather than an embedded process. In this regard, each organisation develops their own methodology according to the stresses of their needs and/or environment, and hence develop their own language and quality standards. Oftentimes these two develop without prior knowledge that there may be a “better” way to do things, as more often than not, the need to document, monitor, advocate, and then share the conditions an organisation finds themselves in come before the ability, and to many the luxury, of premeditated training.

The State of the BPA at the Onset

The BPA exists in this realm: before any external methodology arrived, BPA’s documentation practice had already taken shape under conditions no formal training could have anticipated. In the days following the July 2024 crackdown activists, students, and ordinary bystanders began collecting footage and testimony not as a planned archival exercise but as an act of survival and evidence-preservation happening in real time, often while the events themselves were still unfolding and contributors were still at risk. This is the condition under which most community-led archiving in Bangladesh operates: collection precedes design, and design, when it comes at all, has to be retrofitted onto material that was never gathered with a schema in mind.

Responding to the immense violence these activists, documenters, journalists, and human rights defenders established BPA to preserve evidence for memory, truth, and future accountability. The archive also serves as a foundation for investigations, including through its collaboration with Netra News has meant structuring material to maintain clear chain-of-custody documentation, verifiable sourcing, and exportable case files, so that footage and testimony gathered by community documenters can be used directly in accountability reporting, rather than requiring re-verification from scratch. Their members put themselves into action to gather all they could and make sense of it all. Thousands of videos, pieces of text, images, and witness reports needed to be organised, categorised, and somehow put into a coherent system that could be interpreted, accessed, and understood.

The Volume, Variety, and Quality of Evidentiary Materials was Reaching a Critical Threshold

As the BPA continued to gather evidence and its work became more widely known, the volume of data grew rapidly. At the same time, the archive became a trusted point of contact for protest-related evidence, with journalists, human rights organisations, and other partners requesting access to data or exploring opportunities for collaboration.

This created a two-fold challenge. First, how could we store and structure data in a way that was both systemically organised in a tabular format and suitable for analysis? Second, how could the dataset support the efficient retrieval, interpretation, and sharing of evidence as demand increased?

One might argue the challenge here could be defined according to a logic of scalability and the idea of designing datasets and their storage according to the concept of longevity. In the environment BPA is in, this process by the bare necessity of collecting before it all disappears.

Reorganising such a large volume of data is not a significant undertaking.It requires sustained labour in storage infrastructure, database design, metadata management, and data governance to ensure that evidence is easily searchable and fit for future efforts such as advocacy and investigations.

Methodologies, Guidelines and Handbooks Were Borrowed or Developed from Other Organisations

There is nothing inherently wrong in the borrowing of materials when those materials are intended for adaptation, and certainly there is no harm being done when a small organisation, in the effort to standardise its practise (often a standard set by Western organisations), will develop their documentation according to borrowed material.

The issue is when an organisation borrows and adapts these materials, makes them their own, but in the process of developing them, they do not relate the literature back to a methodological process. 

The key element in developing a methodological process is to develop an instruction manual of the methodological process at the lowest possible level (for example, specific practises for handling specific types of data from specific types of sources, and where they sit in the dataset) that inform the higher processes (for example, we collect videos and we use these columns and values because we want to advocate for X or Y reason). The distinction of these levels are not always defined, and certainly are not obvious.

Unless an organisation releases their low-level methodology to the public, delineating the nuances of data definition and processing, an organisation seeking that knowledge cannot learn the difference. However, even if such a document exists, each and every NGO or researcher is working with unique circumstances, data, and mission objectives. Organisations such as the BPA must be able to produce such literature when the time and energy is available. These materials are often stored in the mind and experience of those working directly with the materials, and must be pulled onto a document and refined.

The way in which this documentation and subsequent refinement happens is, however, contextual. Protest documentation in Bangladesh, and in other countries facing similar unrest, rarely offers respite. The boundaries of incidents blur across days of continuous violence that does not become available immediately. Within the chaos of this, perpetrator identification is harder under mass-mobilised state violence, and sources are frequently anonymous or pseudonymous for their own safety. In addition, protests produce large volumes of footage that is often taken down by platforms before they can be preserved.

The gap is therefore multi-layered: adapting external materials to serve as a way to standardise a precarious methodology, and that any intervention and subsequent creation to a low-level methodology is difficult to achieve when the evidentiary environment is in constant flux.

Understanding that the Audience of your Data is Not the Public

The first question anyone should be able to answer when creating a dataset is: who is this for? For an NGO that question has higher stakes. Who your audience is defines the way you design your datasets.

A key strength for Airwars has always been a strong understanding of this question, defining the evergrowing datasets and these in tandem, granting Airwars the successes it has accrued over the 12 years of its existence. Yes the public should be made aware of the circumstances of the cause you are monitoring, tracking, or the people you are advocating for, but most often than not, it is not the public that has the ability to change the situation. At Airwars we understood that our primary audience is the militaries themselves, and the ministries that govern them. In this light, our datasets have developed in tandem with our conversations to the US-led Coalition, and believe that our success stems from being able to shorten the gap between sharing data and action by working in the format that our audience is familiar with.

In discussions with the BPA it is evident that there are different needs across different realms. There are investigative outlets and journalists; international human rights organisations; victims, survivors, and their families; and the Bangladeshi public itself. Therefore, the database must exist in a way in which it can be used for accountability cases, humanitarian mechanisms, advocacy and legal purposes. And for the public, to function as a record of memory and truth independent of any legal outcome.

These conditions have shaped BPA’s design choices: tiered access, use-specific schemas, and incident-based documentation. Therefore, the conditions for both an audience-centric database design and also a methodology that can effectively serve these audiences is being enacted.

How important this is cannot be understated. Knowing who your data is for shapes the design process, the methodological rigour needed, and the essential documentation needed for a data-dominant organisation.

Knowing who the data is for shapes the existential basis of an organisation: its mission statement, high-level methodological documentation, fundamental low-level documentation and guidelines, and procedures for sharing and analysis. 

Conclusion: Archiving Under Threat

Of course these are observations and thoughts from what is relatively a very short intervention. Each NGO, each activist, each investigator, every person working in similar fields has an extremely specific context that they are tangled with. Processes are developed within this context, and the quality of this development is ultimately the fruit of time and work dedicated to the “meta” of the work organisations like Airwars and the BPA do.

The majority of Airwars is working in environments where their members are not in immediate threat. This can be also said of many organisations that are based in the West that have the luxury to set standards and to develop methodologies. The BPA, like many other NGOs in the region, exists in a constant state of threat.

Their contributors are often simultaneously documenters, witnesses and targets. Creating a constant tension between the pressure to preserve material and the perseverance needed to find, verify and store incidents. Media is constantly at risk of being removed and devices seized. Witnesses and people themselves are always at the risk of being disappeared, persecuted or killed.

Rigorous data lifecycle maintenance, the protection of their sources, and the urgency to collect before it disappears are in recurring conflict. For an organisation working under this kind of pressure, data design is inseparable from the duty of care to the people whose testimony constitutes the archive.

With these considerations in mind, the workshops mark a transition rather than an end point for the BPA.

The immediate goal is moving from an incident-based archive built primarily for preservation, to one organised in such a way that journalists, investigators and such, can use them with no friction in the retrieval process.Although the BPA is already an important contributor of the July Mass Uprising Memorial Museum and other memorialisation initiatives, in the long term, the goal is to further meet evidentiary standards by developing a thorough codebook and data lifecycle methodology, which we believe will make the BPA archive sustainable and available to accountability mechanisms and as a site of memorialisation for years to come. 

The data design work done through this process is what makes that possible: without it, BPA would have remained a large, well-intentioned, but scattered collection that each new investigation had to re-verify from the ground up; with it, the archive itself becomes reusable infrastructure rather than a one-time record.

Photo attribution: A Chosen Soul on Unsplash