Credit: Lieber Institute of West Point
by Edward Madziwa, Program Assistant for the Security and Technology Program at the United Nations Institute for Disarmament Research
The loss of life is an undeniable and tragic reality affecting all parties to armed conflicts raging across the world. Thousands of unidentified victims are left in a state of decomposition. In some instances, corpses are mutilated, desecrated, abandoned without any funeral rites, or not afforded decent burials. International humanitarian law (IHL) seeks to limit the means and methods of warfare and to protect war victims. Parties to armed conflicts are bound to collect and search for the dead, to prevent the dead from being despoiled or mutilated, to ensure that the disposal of the dead is done in a dignified manner by following the deceased’s culture and religion, and to account for the dead. In international armed conflicts (IAC), there is an obligation to return the human remains and personal effects when possible.
With technological developments presenting tremendous opportunities, rapid developments in data-driven artificial intelligence (AI) and machine learning (ML) have the potential to significantly transform humanitarian forensic action. However, their role in forensic identification of the deceased remains unexamined. Over the years, tech enthusiasts in collaboration with forensic practitioners have developed tools like Skeleton ID, Commingled Remains Analytics (CoRA), and forensic facial imaging tools to aid in the identification of the missing and the dead.
The application of AI and ML in humanitarian contexts can revolutionize forensic practices, allowing for the integration of various identification techniques while increasing the speed and accuracy of matching remains. Using a multidisciplinary approach, this post explores AI and ML challenges and opportunities in advancing honour and dignity in death for armed conflict victims in humanitarian forensic action under IHL.
The Use of AI and ML in Forensic Action
AI and ML have been integrated into civil protection activities to reduce disaster-related deaths, identify vulnerabilities, detect hazards, and predict future risks. They were applied in the context of the COVID-19 pandemic for faster diagnosis and triage by healthcare professionals. Infervision Technology has deployed AI-based software for infrared computerized tomography scans, detecting pneumonia lesions with an 83% accuracy rate. Chinese facial recognition firms like SenseTime and Megvii have modified their AI systems to identify people wearing masks and those exhibiting coronavirus symptoms. In armed conflict situations, AI and ML are extensively used in geospatial intelligence, analysing open-source material in sensitive geopolitical locations and predicting and locating mass graves in post-conflict situations.
Current Trends in and Use of AI and ML in Humanitarian Forensic Action
Human identification is crucial in humanitarian forensic action and medico-legal death investigations. Popular methods include visual recognition and fingerprint matching, DNA testing, and comparative dental analysis, but these require prior records and preserved materials for comparison and analysis. In cases with poor preservation or bare bones, identification becomes difficult. Advances in ML have improved image recognition, restoration, and synthesis, aiding forensic practitioners in parameter estimation, image segmentation, pathology classification, and image enhancement.
The Korean War Identification Project
The 1953 Korean Armistice Agreement allowed for the recovery and exchange of human remains between opposing forces during the Korean War. However, thousands of dead U.S. soldiers could not be handed over before the complete closure of the border between North and South Korea. This delay led to commingled remains that inhibited recovery. In response, the United States developed predictive models and full-spectrum identification algorithms to compare isotopic values in tissues, enabling detection systems to adapt to diverse radiation fields. Software programs like CoRA also help separate commingled remains into distinct individuals.
Forensic Facial Imaging
ML algorithms and forensic facial imaging software have been developed to aid in human identification, including facial approximation and photographic superimposition. Face Lab, a research group at Liverpool John Moores University, focuses on facial identification using forensic facial imaging. Their research includes developing a 3-D computerized craniofacial representation system, a database of anatomical traits, and advanced facial depiction, animation, and recognition technologies, all of which facilitate humanitarian forensic action.
Skeleton ID
Skeleton ID uses physical anthropological methods like craniofacial photographic superimposition, biological profiling, and comparative radiography for skeleton-based identification. While advances in AI and ML are welcome, the nature of these techniques makes them only useful in limited circumstances. Where there are widespread bombings with fragmented remains, in the absence of skeletal data, they are bound to fail. Multiple lines of evidence, including dental, genetic, medical, contextual, geographical-temporal, biological profiles, and individualizing features, can be used for identification.
IHL Treaty Law Obligations Relevant to Humanitarian Forensic Action
The Geneva Conventions and other IHL rules impose obligations on warring parties to preserve the dignity and humanity of the deceased and their families. These obligations apply to all armed conflict situations, including non-international armed conflicts (NIAC).
The Obligation to Collect and Search for the Dead
The duty to search for and collect the dead is a long-standing IHL rule codified in the 1949 Geneva Conventions. It requires parties to take all possible measures to search for, collect, and evacuate the dead whenever circumstances permit, without any adverse distinction on any basis. The first Additional Protocol (AP I) urges warring parties to agree on arrangements for teams to search for, identify, and recover the dead, and to allow civilians or humanitarian organizations to assist.
In NIACs, the conflict parties bear the obligation to search for the dead and missing. Certain obligations within the second Additional Protocol (AP II) only apply if the non-State party to the conflict exercises territorial control. However, IHL obligations are not reciprocal, and parties must comply with their obligations regardless of whether they apply to others.
The Disposal of the Dead
Parties to a conflict must ensure dignified burials after a medical examination to establish identity. This obligation was codified in the 1929 and 1949 Geneva Conventions. AP I mandates parties agree to permanently protect and maintain gravesites. Burials must follow the deceased’s religious rites. Cremation is recognized as an exception for hygienic or religious reasons. Warring parties must ensure individual burials, avoiding collective burials without prior identification unless urgently necessary. Graves must be respected and organized according to the deceased’s nationality.
Humane Treatment
The requirement for humane treatment is outlined in the Geneva Conventions, the Additional Protocols, and amounts to customary IHL. At a minimum, the requirement for humane treatment includes the prohibition against the mutilation of dead bodies and taking measures to prevent the despoiling of bodies. Mutilation of dead bodies is considered a criminal violation of IHL. This prohibition upholds the dignity of the deceased and demonstrates respect for their relatives.
In some armed conflicts, gruesome images and videos of dead soldiers have been posted on social media platforms, accompanied by messages encouraging families to verify whether the images depicted their loved ones. While the objective was to facilitate identification, the method employed could amount to exposure to public curiosity, potentially violating humane treatment obligations under the Geneva Conventions. The concept of “exposure to public curiosity” under the third Geneva Convention applies to prisoners of war. However, the same concept also applies to the families of the dead.
Return of Remains
In IACs, the Geneva Conventions do not establish clear obligations to return the deceased’s remains. However, Articles 34(2) and (3) of AP I suggest a proactive approach, while Article 17 of GC I allows for the return of remains if circumstances allow. Article 120(6) of GC III acknowledges the importance of respecting families’ wishes, and Article 130(2) of GC IV provides for the return of ashes “as soon as possible” to the next of kin on request. Several United Nations and Red Cross and Red Crescent Movement resolutions urge belligerent parties to facilitate the repatriation of the deceased’s remains.
In NIACs, the applicable IHL treaty provisions are silent on the return of remains from one belligerent party to another. However, facilitative agreements show the possibility of returning dead bodies in NIACs, as long as it is in accordance with the requirement of respect for family life.
Customary IHL
Customary IHL rules on the dead address gaps in the law governing both IACs and NIACs. The International Committee of the Red Cross (ICRC) opines that all rules regarding the dead apply in both types of conflict, except for the rule on the return of the deceased’s remains and personal effects. The ICRC Customary IHL Study suggests that this rule should apply equally to both IACs and NIACs, requiring States to take a proactive and positive stance.
Opportunities for AI and ML in Humanitarian Forensic Action=
There are various opportunities and challenges that AI and ML present to humanitarian forensic action.
AI and ML can automate repetitive tasks like image recognition, image and video analysis, and large-scale data analysis, reducing backlogs and allowing experts to focus on data analysis and identifying the deceased’s remains. This can lead to timely identification and minimize delays in returning remains to families for burial or memorialization.
AI and ML can accelerate the forensic analysis process by modelling and structuring human experts’ knowledge and extracting new knowledge from vast databases. They can process and analyse enormous datasets, providing excellent tools for resolving complex datasets. They can determine independent complicated facial recognition features or parameters for challenging tasks. AI and ML can help reduce timeframes by analysing large datasets and facilitating identifications in a more timely manner. This can help families of victims of armed conflicts honour their loved ones by providing a befitting burial in accordance with their religious and cultural beliefs and promoting emotional healing, a fundamental aspect of the right to dignity.
AI and ML can enhance human identification accuracy by providing efficient pattern recognition in forensic science. These technologies, based on statistics and probabilistic reasoning, can identify complex patterns in data collected before death, potentially eliminating human subjectivity and errors while reducing false positives or negatives.
Challenges for AI and ML in Humanitarian Forensic Action
Right to Privacy, Data Vulnerability, and Privacy Concerns
The right to privacy is crucial for safeguarding human dignity and ensuring the safety of data subjects. Data protection laws define “data subjects” as natural persons who can be identified directly or indirectly by reference to personal data. Consent from the deceased family is essential for data handling, but other legal bases may be used in cases where the deceased’s family is unknown, deceased, or estranged from them. While individuals may consent to data processing for humanitarian action-related purposes, they may not be aware that their data may be used to develop AI systems. This data may be vulnerable to theft, abuse, and destruction, particularly in armed conflict contexts. Indefinite data retention may expose it to theft, destruction, or abuse.
Bias and Discrimination
AI and ML algorithms can exacerbate existing biases and discrimination. The 2018 Toronto Declaration aims to make human rights a key component of data ethics in ML systems. The context in which an algorithm is deployed and the task it is supposed to perform determine whether it is discriminatory. Developers of AI and ML technologies for humanitarian forensic action must be conscious of the diverse human rights applicable in this context. Failure to do so can lead to disproportionate misidentifications and emotional distress for families who may have received inaccurate information about their loved ones.
Dehumanisation
While much is made of the new capabilities offered by AI and ML, a realistic assessment of these technologies’ capabilities and limits is essential where they are used for armed conflict-related activities. These technologies, while advanced, cannot truly embody human traits. Given the risks inherent in conflict zones, the deployment of AI-enabled humanoid robots to search for and collect the dead may be an option. However, since AI-enabled humanoid robots lack genuine emotions and empathy, their use may not adequately address diverse cultural practices surrounding death and mourning and may have a dehumanizing effect on the deceased and their families. They may also struggle to recover entire remains or to treat commingled remains with dignity. Therefore, their use should be limited to identification tasks, not the collection of the deceased’s remains.
AI and ML systems should be used to supplement human experts, not replace them. Human experts are capable of making value-based decisions that uphold broader ethical standards, which may not be incorporated into AI or ML systems. Therefore, human oversight is crucial over tasks that undermine the deceased’s dignity and honour.
Asymmetry
The use of AI and ML in humanitarian forensic action could potentially highlight asymmetry between warring parties. While these technologies can help to fulfil the parties’ IHL obligations, not all parties have the same technological and financial means to use them. The costs and feasibility of different identification methods are more pronounced in conflicts with varying military power and resources.
While IHL recognizes the equality of belligerents, non-State armed groups in NIACs may not use these technologies to fulfil their humanitarian forensic action obligations. This could undermine fundamental IHL principles, particularly the principle of humanity, as non-State armed groups lack the finances, technology, or willingness to employ costly forensic methods.
Recommendations
AI and ML can be used in humanitarian forensic action to advance the dignity of armed conflict victims. However, there are drawbacks. To maximize the opportunities offered by AI and ML in humanitarian forensic action, States should adopt legal, operational, and policy interventions like those mentioned below.
Adopt a Binding International Instrument Governing the Use of AI and ML
Regional efforts to regulate AI are commendable, but a comprehensive international instrument with a model law would ensure uniformity and allow for regional modifications based on culture, background, and history.
Develop Guidelines on the Responsible and Ethical Use of AI and ML Technologies
Experts should develop guidelines that would assist in the incorporation of AI and ML into humanitarian forensic activities. These guidelines should align with the principles enshrined in IHL and international human rights law, prioritizing the “do no harm” principle and ensuring that AI and ML do not cause additional trauma or distress to those affected.
Training of Humanitarian Forensic Actors
The ICRC could collaborate with other humanitarian organizations to provide training programs on IHL and human rights law principles. Such programmes could include training on how AI and ML technologies can assist in humanitarian forensic action and the potential of aligning them with IHL and human rights law principles. This would help bridge the gap between technology and humanitarianism.
Additionally, training should be extended to developers of AI and ML technologies for humanitarian forensic action, preventing discriminatory trends. This training would enable developers to develop oversight mechanisms to identify, examine, resolve, and test biases in datasets and models during the designing and development phases.
Conclusion
AI and ML can revolutionize human remains identification and ensure dignified burials in mass fatalities caused by armed conflict. These technologies can speed up the process and provide families with more information about their loved ones’ fate, potentially aiding in the healing process. However, skilled forensic experts, anthropologists, and legal professionals are essential.
IHL and human rights standards should guide the deployment of AI and ML in humanitarian forensic action. Integrating AI and ML can advance honour and dignity in death for victims of armed conflicts, ensuring dignity for victims and their families through ethical concerns and a human-centred approach.
About the author
Edward Madziwa is a Programme Assistant for the Security and Technology Programme at the United Nations Institute for Disarmament Research (UNIDIR) in Geneva. This excerpt is drawn from the author’s article-length work, “Advancing Honour and Dignity in Death for Victims of Armed Conflicts: Exploring the Challenges and Opportunities of AI and Machine Learning in Humanitarian Forensic Action Under IHL” appearing in the International Review of the Red Cross. Republished courtesy of Lieber Institute of West Point