Quick Facts Contact

About GReVD

The Consortium

Resources
a

About GReVD

In this Section:

Challenges

Key:

Done

In Progress

Next

On the Challenges page, the Consortium tracks recent accomplishments, current research priorities and scans the horizon for obstacles that will need to be overcome to build the Registry.  

Users can toggle short descriptions for each challenge (show/hide) by using links below.  

show all descriptions|hide all descriptions

Immediate

Coding

Precision

Registry

Consortium Support

Challenges of coding violent deaths indentified in the Gaps Report are summarized in this four page brief.  Five channels for coding violent deaths are shown with issues for under and over counting at nine stages of coding.  

Brief 3 (PDF) f

Issues with defining violent deaths were identified in the Gaps Report (forthcoming) and are summarized in this brief which asks the question, "What is a violent death?"

Brief 2 (PDF) f

This brief describes the objectives and priorities for GReVD as well as an overview of the consortium and member partners.  

Brief 1 (PDF) f

In a quickly changing world, the GReVD consortium is looking at the relationship between COVID-19 and conflict risk. The disease—or measures intended to contain it—could conceivably increase ethno-sectarian tensions, intensify political repression and unrest, or undermine the economic conditions that underpin stable political systems.  The social effects of the disease are unlikely to end when infections peak—they may be felt for years to come.

Existing data collection efforts are unlikely to capture these causal processes quickly enough to enable effective intervention and policy analysis. There are gaps in our measurement of lethal violence, as well as and non-violent but disruptive events like supply chain failures, food shortages, and trade disputes

The GReVD consortium has documented technical challenges to producing timely, accurate data on lethal violence around the world, as well as innovative technologies and tools that could accelerate the production of policy-relevant data. This evidence will be necessary to assess interventions that work to sever the link between contagion and conflict.

CN for AI Application f

The COVID pandemic demonstrates the importance of good evidence and data for policymaking.   

The GReVD Consortium is involved in discussions about global information management.  Information management and the governance necessary to create and use global statistics.  

If you are working on issues related to solving these challenges, please contact the GReVD consortium through the contact us page.  

The Earthtime platform at CMU will be used for visualizing the GReVD data.  

Three initial visualizations will be produced highlighting some of the key challenges in counting violent deaths, as detailed in the Gaps report. The first one will visualize for each country where data exist, the gap between the upper- and lower bound estimations of violent deaths. The second will showcase the challenges of integrating datasets by visualizing how the same violent death may be coded differently across existing databases. The third visualization will demonstrate how much missing information there is, globally, in the count of violent deaths, by mapping “fog of war” – where data is incomplete or missing.

The first workshop was held in February 2019 in two parts (Washington DC and Geneva) at Brookings and SAS (respectively).  This workshop was formative in the resulting Gaps Report. 

The second workshop was in Washington DC, 31 October 2019 at Brookings.  The workshop was titled "Robots and Ghosts" as the main topics were machine coding and "known unknowns" of missing information for known incidents of violent deaths.  

Cline and SIPRI collaborated on two thinkpieces early in the GReVD initiative as background work on the current state of machine coding of violent deaths.  The first of these pieces was a technical paper describing  coding processes and issues and obstacles that arose.  The second is a memo with some thoughts on steps forward.   These are available from the consortium by request.  

The gaps report is a stocktaking of current coding practices and an assessment of challenges with creating a Global Registry. 

A preliminary (draft) version of the Gaps Report is available from the consortium by request.  A final edited version of the report will be added here once complete.  

The consortium endorsed the Global Violent Deaths (GVD) data produced by Small Arms Survey as the lower bound (conservative) estimate of violent deaths.  

This document provides a review of the GVD approach and the complementarity of the GVD with GReVD.  

Brief 4 (PDF) f

Build the basic website to convey information about the consortium, partners, current challenges (research priorities), research outputs and principles and standards.  

Pilot "Machine in the Loop" consultation with GTD and Cline, overseen by SAS.  Report forthcoming.

Short paper on ICCS, DC-10 and other coding rules (including references for relevant coding rules from consortium partners).

Design database structure drawing on raw violent death event data from ACLED, GTD and UCDP.

This first iteration does NOT include deduplication yet.  It is not a substitute for GVD estimates of violent deaths.

Convene 3rd Workshop on Machines and Human Coding

Learn from current coding processes at ACLED and UCDP (3rd Workshop?)

Document current state of knowledge on human coding and prospects for machine contributions

Administrative data on homicides can come from a variety of sources.  This small research initiative will compare (quality, levels and trends) in provincial level data with country level data compiled by Igarape through the Homicide Monitor.  

Build Registry v0

Launch of first version of Global Registry of Violent Deaths

Identify roles for machines in the loop. Identify future priorities for machine reading and machine coding.

Estimate “known unknowns” and introduce the concept of "fog" - a measurable level of uncertainty based on quality of administrative statistics (measures of statistical capacity), freedom of the press and other relevant metrics. 

Identify Use Cases for Registry

Regular reporting on lower and upper bound estimates of violent deaths

Transfer current Violent Death data from ACLED, UCDP, GTD

Define known unknowns in Registry, allow for placeholders.

Fundraising for mini-projects and proof of concepts for future challenges.

Consortium definition of violent deaths, shared ontology

Design backend that connects event records from ACLED, UCDP and GTD to Registry

Develop single estimate of violent deaths, using upper, lower bound estimates and known unknowns

Medium Term

Coding

Precision

Registry

Consortium Support

Engage with health experts on admin data (WHO) to connect ontology and definitions.

Necessary for statistical capacity building with national actors.

Develop 1st front end for Registry using Earthtime visualizations

Wireframe for Registry interface

Engage with criminal justice experts on admin data (UNODC) to connect ontology and definitions.

Largely related to issues of intentionality in coding of violent deaths.

Necessary precondition for statistical capacity building with national actors.

Monitor known unknowns and fog over time to demonstrate impact of GReVD

Implement machine reading

Iterative development of validation processes in Registry (validation, disambiguation, deduplication)

Long Term

Coding

Precision

Registry

Consortium Support

Allow for filtering of data in Registry by ontology

Backend interface that connects administrative data to Registry

As contributor base increases, a reputation system (adapted from crowdsourcing approaches) will be necessary to monitor quality of contributions by individual coding platforms.  

Statistical Capacity Building for pilot countries to improve administrative data coding

Ongoing State-Society discussions of Violent Deaths, coding and monitoring, using Registry

Implement machine coding

Develop standards and practices for crowdsourcing in registry

Implement blockchain or other secure ledger processes to ensure integrity of Registry, attribution and compensation for coding.

Backdating and counting of violent deaths using expanded corpus

Historical Registry

2030

Coding

Precision

Registry

Consortium Support

Statistical Capacity Building for developing countries to improve administrative data coding

Reporting for 2030 progress on reducing Violent Deaths

Machine coding for Administrative Data Sources (criminal justice documents, court reporting, hospital records)