Technology is not neutral

First, humanitarians need to explicitly acknowledge and integrate the fact that innovation and technology are neither neutral nor necessarily good in and of themselves.Footnote ³⁵ They constitute complex socio-technical constructs carrying underlying but significant assumptions and values, usually aligned with those of the people who develop and promote them.Footnote ³⁶ The “digital transformation” is, for instance, not a mere factual description of a trend, but also an agenda for change that comes with implicit but important structural shifts reflecting neoliberal and capitalist orientations.Footnote ³⁷ It is shaping and shaped by society and by political and economic interests, and it triggers serious consequences and impacts.Footnote ³⁸ Neglecting these considerations comes with significant risks for humanitarian action.Footnote ³⁹

Academics have already highlighted how these structural shifts are impacting the humanitarian sector. Its centre of gravity is moving from a focus on mostly physical and human methods to the technological and digital, from largely public and non-profit to hybrid and commercial approaches, and from a central role of States and governments to the growing role of private sector actors.Footnote ⁴⁰ Humanitarian organizations are recruiting more data scientists and AI experts, but fewer anthropologists and ethnologists. They are developing public–private partnerships, but are less at ease with civil society organizations. Their donors are requesting more “value for money”,Footnote ⁴¹ but not necessarily human rights impact assessments.Footnote ⁴²

The consequences of those shifts are contributing to what has been described as a “privatization”,Footnote ⁴³ “commodification”Footnote ⁴⁴ and “marketization”Footnote ⁴⁵ of the sector. This evolution is reflected in the changing vocabulary of humanitarian professionals.Footnote ⁴⁶ It has become relatively common to hear discussions on improving “productivity” (instead of impact) through innovation and ensuring “scalability” (instead of relevance). People affected by conflict and other humanitarian crises have over time turned into “customers” and “clients” for humanitarian “services”.Footnote ⁴⁷ Donors request “returns on investment” and private sector partners offer expertise in leveraging market opportunities for “social good”.Footnote ⁴⁸ Analyzing this lexicon is interesting because it reveals the assumptions and dynamics that come with it – namely, a focus on perceived efficiency and measurable outputs at the cost of qualitative humanitarian outcomes, and on market-based commercial strategies at the cost of needs-based humanitarian approaches.

Humanitarians have a lot to learn from private sector actors in terms of efficiency and the ability to deliver on commitments. Partnering with private companies can help improve effectiveness and management practices, and humanitarian organizations have been notoriously and legitimately criticized for their failures in these domains.Footnote ⁴⁹ But these partnerships do not always work two ways, and humanitarian actors are often more impacted by the transfer of knowledge and values from the private sector than the other way around. This is particularly true in the field of innovation and new technologies.Footnote ⁵⁰

It is argued that “while ‘technology’ and ‘the private sector’ have both been constant entities in the humanitarian sector”, their significant influence in humanitarian innovation “represents something qualitatively new”,Footnote ⁵¹ “changing the very nature of humanitarianism”.Footnote ⁵² The assumptions integrated in tech companies’ business strategies and practices can have a transformative impact on humanitarian ethics and practice.Footnote ⁵³ Private tech companies’ utilitarian approach to humanitarian partnerships is understandable because the partnerships represent good branding, visibility, corporate social responsibility and new market entry possibilities, among other incentives.Footnote ⁵⁴ This approach, however, comes with supply-driven opportunistic and experimental methodologiesFootnote ⁵⁵ that do not necessarily align with the needs-driven and precautionary ones that ought to characterize humanitarian practices.Footnote ⁵⁶

“Techno-solutionism” and utilitarian approaches

Tech companies, by their purpose and nature, function on the business- and profit-driven assumption that technological innovations are intrinsically good and can fix all sorts of human and societal problems. This so-called “techno-solutionism”Footnote ⁵⁷ is difficult to resist for humanitarian organizations faced with unbearable levels of suffering, intractable needs and limited resources in particularly fluid and insecure environments. There is therefore high humanitarian receptivity for “tech-utopianism”Footnote ⁵⁸ and opportunistic solutions to alleviate suffering.Footnote ⁵⁹ In simple terms, the prevailing attitude is: if tech can help, let's use it.

This pragmatism is explicit in the humanitarian narrative vis-à-vis emerging technologies, which focuses on leveraging the opportunities they create while mitigating the risks they bring, in order to help “more” people.Footnote ⁶⁰ This approach seems a priori adapted to the double-edged nature of digital technologies. Utilitarian ethics are neither unfamiliar nor illegitimate in humanitarian action: prioritizing solutions that help improve the situation of as many people as possible makes sense. The problem is that utilitarian ethics alone are insufficient in the humanitarian context.Footnote ⁶¹ Deontological, value-based and professional ethics are important complementary guard-rails against overly pragmatic choices – and this is where the humanitarian principles, taken as an interdependent whole and in hierarchical order, are useful checks and balances.Footnote ⁶²

Indeed, a more granular analysis of the “opportunities versus risks” utilitarian frame based on the techno-solutionist promises and assumptions of tech companies often reveals that the binary equation is not a fair one. In practice, the supply-driven opportunities that these technologies offer can be “solutions in need of a problem” for which humanitarian action constitutes an interesting testing ground.Footnote ⁶³ This framing can reverse the humanitarian problem identification process from a problem-driven approach to one driven by solutions.Footnote ⁶⁴ Instead of asking if and how new technologies can help alleviate suffering, the leading question becomes if and how humanitarian needs can help “keep up” with new technologies. Instead of being a means to achieve humanitarian ends, the use of new technologies becomes the end, and humanitarian needs the means to achieve it. Instead of being “bottom-up” and triggered by operational needs and challenges, innovation becomes “top-down” and justified by considerations of competitive relevance and the availability of technology and funding.Footnote ⁶⁵

As a result, technological choices are often used to address the problems of humanitarian organizations, their donors and their partners. For example, digital advances have improved traceability for fraud prevention and security by improving the identification and authentication of people in need – but this process can sometimes create or reinforce a de facto presumption that affected populations are potential fraudsters or security threats, highlighting a reversed burden of proof and a lack of trust in them.Footnote ⁶⁶ New technologies have also improved scalability, creating economies of scale in operational delivery methodsFootnote ⁶⁷ and enabling bureaucratic cost reduction for humanitarian organizations – but this process can lead humanitarians to neglect the differences and specificities of individual contexts. These dynamics can reverse humanitarian logic: while the opportunities are mostly benefiting the providers of humanitarian aid and their partners, the risks are mostly carried by the populations at the receiving end, in particular when the personal data they provide to feed innovative digital solutions are not adequately protected.

Any humanitarian interventions, innovative or not, are likely to cause some degree of harm.Footnote ⁶⁸ Utilitarian ethics illustrate this reality in requesting a positive balance between that level of consequential harm and the greater good achieved. What is fundamental is that this calculus is made explicit, adequately assessed, and accounted for to the extent possible. In the context of humanitarian innovation, those requirements often seem to have become neglected afterthoughts.Footnote ⁶⁹ Despite efforts to improve accountability to affected populations and increase their participation in the design and delivery of humanitarian responses,Footnote ⁷⁰ affected people still do not really contribute to decision-making or risk analysis attached to the specific deployment or use of innovative technological solutions by humanitarian organizations.Footnote ⁷¹ When they are consulted, it is often to support a confirmation bias, and without allowing or helping them to truly understand what is at stake.

The attached risks are therefore imposed on them, sometimes without their knowledge or truly informed consent – including when those are required due to the processing of sometimes sensitive personal information.Footnote ⁷² It is often argued that, when asked, affected people would certainly want to have access to innovative digital tools and that it would be paternalistic to deny it to them. Like everyone else, affected people have a fascination bias towards technology, and a desire not to be left behind in the wake of the digital transformation. But this does not mean that they understand what the risks attached to digital technologies are in their contexts, or that many of the safeguard mechanisms available to others are necessarily available or functioning where they find themselves. To better respect their safety, dignity and autonomy, humanitarians have the duty to go beyond this assumption and to help them be in a better position to make a truly informed decision. This requires explaining to them, in a language they understand, why a specific technology is used for a particular programme, and translating into real-life examples what the risks and consequences of using that technology may be.

In many situations, however, affected people are not even asked for their opinion, and this is also true for the use of innovative technology for humanitarian action. It is often assumed that in their dire situation and exceptional circumstances, the risks attached to the digitalization of humanitarian responses are justified by the potential gains.Footnote ⁷³ In the context of humanitarian emergencies, suffering and urgency are sometimes used as excuses to justify experimentation, exceptionality and higher risk appetite at the cost of safety and ethical guaranteesFootnote ⁷⁴ that can be seen as obstacles to action and immediacy. This trade-off management approach is ethically questionable because it contradicts the “do no harm” requirement attached to the humanitarian principle of humanityFootnote ⁷⁵ and does not respect the agency that is so central to the dignity of affected people. When technological experimentation can cause or lead to real additional human harm, it creates a risk of defeating the very purpose of humanitarian action: the alleviation of suffering. And when new technologies are deployed without the participation or consent of affected people, they neglect those people's agency and ability to participate in decisions that affect their lives. These potential drawbacks are significant pressure points on the ability of humanitarian actors to operate in line with the principle of humanity.

Generative AI and the risk of degenerative humanitarianism

One of the key tenets of the humanity principle resides in the sentiment or attitude of someone who shows themself to be human.Footnote ⁸⁰ Yet, one of the objectives of the digital transformation – particularly with AI – is to use technology to perform tasks normally carried out by humans. In a sense, it aims to “de-humanize” certain activities. As a result of digitalization, some humanitarian activities and processes are likely to become literally less human because the professional “aid deliverer”, or the interface that represents it, becomes a machine rather than a human being. In some domains – such as information or financial management – this transformation is not necessarily problematic, and can reduce the burden attached to repetitive and unpleasant but necessary bureaucratic tasks.Footnote ⁸¹ In others, where empathy is important, it raises significant questions related to the increasing disappearance of humans, and their ability to demonstrate empathy and understanding, in the delivery and management of humanitarian activities.

Respecting the dignity of people in need implies the ability to show empathy and to understand their situation or feelings. This requires an ability to listen and to discern the complexities and nuances of their experiences, as dignity is a personal feeling that is necessarily self-defined.Footnote ⁸² This explains why humanitarians have always attached importance to being physically present where affected populations find themselves. When humans are replaced by digital interfaces that introduce different forms of intermediation and remoteness – for instance, when a trained humanitarian worker able to show empathy is replaced by a smartphone app for “self-registering” needs – this essential proximity element is mechanically lessened.Footnote ⁸³ It is therefore important that humanitarians strive to use digital tools to enhance, and not replace, human interactions – but it is essential to understand that their efforts to do so will be jeopardized by the pervasive nature of digital technologies, which tend to spread and expand organically.Footnote ⁸⁴

An illustrative example of the digitalization of human interactions is the development of “humanitarian chatbots”. It is argued that “in recent years, chatbots have offered humanitarian operators the possibility to automate personalised engagement and support, inform tailored programme design and gather and share information at a large scale”.Footnote ⁸⁵ According to the Office of the UN High Commissioner for Refugees (UNHCR), chatbots

Without entering into a detailed analysis of the pros and cons of these tools – which others have aptly analyzedFootnote ⁸⁷ – the above statements confirm that these innovations are mostly geared towards organizational interests (i.e., data collection and scalability) and rely on the assumption, or confusion, that humans and machines are interchangeable.

But research is showing that they are not. Indeed, humans do have a natural tendency to anthropomorphize technological innovations and give them human attributes that they do not possess. The so-called “Eliza effect”Footnote ⁸⁸ is a cognitive bias associated with textual interface computer programs, leading users to believe that the machine has human capabilities such as intelligence or empathy. Impressive progress in machine learning and large language models (LLMs) in recent years has considerably improved chatbots’ performance and has magnified this cognitive bias (on which AI marketing lexicon and visuals heavily relies),Footnote ⁸⁹ leading to debates on whether these machines can be “sentient”Footnote ⁹⁰ or perform “any” human tasks (i.e., “artificial general intelligence”).Footnote ⁹¹

A recent scientific study demonstrated that in the health-care domain, patients tend to prefer the quality and empathy of chatbots’ responses to their medical questions compared to the responses of physicians.Footnote ⁹² This finding highlights the power of the AI systems behind chatbots and their ability to generate an illusionary feeling of empathy in their users. Yet, clinical experts have documented the negative secondary impact of artificial empathy in terms of trust and effective care, pointing to the impossibility of truly replacing human empathy with an AI version of it.Footnote ⁹³ According to them, the illusion is eventually counterproductive, and it is critical to maintain “human monitoring and emotional intervention” in order to ensure effective empathy. Ignoring that requirement can trigger “difficult moral and legal responsibility” issues for medical professionals.Footnote ⁹⁴

This lesson is highly relevant for humanitarians, for whom the concepts of care, trust and effectiveness are indispensable. The negative long-term secondary effects that chatbots may have on the effectiveness of humanitarian care and the trust of the populations who now partly depend on these technologies put at stake the human elements without which the humanity principle cannot be operationalized. Care and trust are not productivity metrics, but essential requirements of the humanitarian endeavour. Without appropriate safeguards, the growing use of chatbots and other “generative AI” systems risks becoming the seed of degenerative humanitarianism.

Protecting data is protecting people

Digitalization brings other challenges vis-à-vis the principle of humanity. Some have argued that the process of “datafication” – which turns information about people into data points feeding “quantitative processes that used to be experienced qualitatively”Footnote ⁹⁵ – is itself a dehumanization process necessarily reducing people's complex realities and identities to intangible data and predefined categories.Footnote ⁹⁶ It seems indeed questionable that the binary nature of data,Footnote ⁹⁷ or the categorization labelling requirements of databases and LLMs, can adequately capture the complex and multifaceted identities, experiences and needs of people affected by humanitarian crises.Footnote ⁹⁸ Instead of making data fit for humanitarian needs, the focus is on fitting needs into data and getting as much data as possible about them.Footnote ⁹⁹ Unstructured, localized, nuanced and qualitative information are not in sync with the data hygiene requirements needed to facilitate data flows.Footnote ¹⁰⁰ “Datafication” can therefore lead to “ignor[ing] or even smother[ing] the unquantifiable, immeasurable, ineffable parts of human experience”Footnote ¹⁰¹ and can reduce affected people to their “electronic double”, or mere digital avatars.Footnote ¹⁰² When these important nuances are lost to quantitative methodologies supporting the auditing and securitization objectives of digital innovation,Footnote ¹⁰³ there is a significant risk that those methodologies will end up replacing qualitative ones, eventually eroding the quality of humanitarian responses.

Other concerning threats hide behind the growing number of humanitarian data flows. Humanitarian data streams contribute to understanding and documenting needs or international humanitarian and human rights law violations, but they can also “be a causal vector for harm” through “unpredictable or unpredicted knock-on effects”.Footnote ¹⁰⁴ For example, if such data are not adequately protected, personally and demographically identifiable information – combined with other, less sensitive data through a “mosaic effect”Footnote ¹⁰⁵ – can help identify and locate people for surveillance or targeting.Footnote ¹⁰⁶ Such data can also help to draw up their political or emotional profiles for influence or manipulation purposes.Footnote ¹⁰⁷ In many countries and for most people, the mis- or repurposed use of these data usually leads to targeted online advertisements and unsolicited commercial offers – what has been described as “surveillance capitalism”.Footnote ¹⁰⁸ In conflict and other violent settings, it may expose vulnerable people or communities to a different set of risks and can lead to targeted killings, arrests, persecutions or disinformation and manipulation. As recalled by former US Central Intelligence Agency director General Michael Hayden, security agencies use seemingly innocuous metadata to kill people.Footnote ¹⁰⁹ While is it virtually impossible to know if and when humanitarian data are used to target or persecute people, the absence of evidence should not be taken as evidence of absence, and the risk needs to be taken seriously.

The humanitarian sector has made enormous progress in recent years in acknowledging the importance of personal and humanitarian data protection,Footnote ¹¹⁰ and in minimizing the potential harms triggered by the sector's digital activities, in line with the so-called “do no digital harm” concept.Footnote ¹¹¹ Most organizations have adopted data security policies and guidelines, hired experts and set up data protection offices. Yet, these tools are inspired by data protection rules and precepts – in particular the European Union (EU) General Data Protection Regulation (GDPR)Footnote ¹¹² – created by and for countries with significant legal and technical means, and even in those countries, their implementation remains an enormous challenge.Footnote ¹¹³ In places affected by conflict and disasters, where institutional safeguards and control mechanisms (such as data protection legislations or controlling authorities) are often dysfunctional or non-existent,Footnote ¹¹⁴ implementing data protection standards becomes virtually impossible.

Ensuring effective data protection requires informing “data subjects” about why and how their data is being collected, used or shared. Humanitarian organizations face significant transparency and accountability challenges with regard to their increasing data collection reliance and helping affected population understand the related trade-offs.Footnote ¹¹⁵ It requires time and effort to translate complex technical issues into a language that people understand; it also implies breaking the power imbalances that characterize aid provision and offering true alternatives to the data-for-aid bargain now so often implicitly embedded in digital humanitarian processes.Footnote ¹¹⁶ Yet, in the context of emergency humanitarian responses, these important requirements can sometimes be considered as burdensome obstacles to speed and scale, or disregarded based on certain assumptions.Footnote ¹¹⁷ As a result, informed consent becomes an afterthought, increasingly neglectedFootnote ¹¹⁸ in favour of other data collection bases such as the “legitimate interests” of the organizations who collect the data – sometimes behind the shield of legal privileges and immunities, which are important safeguards that should not be misused.Footnote ¹¹⁹

As most people have experienced, clicking on the “I agree” button to accept the lengthy and complex “terms and conditions” applying to digital services without understanding them is not exactly a satisfying experience of informed consent, control or agency.Footnote ¹²⁰ There are valid concerns that the “informed consent” concept is no longer fit for purpose in a digitalized world controlled by asymmetrically powerful tech companies.Footnote ¹²¹ However, by neglecting informed consent in their practices, humanitarians are imposing risks on affected populations without having adequate means to be responsible or accountable for their potential consequences.Footnote ¹²² This effectively turns “data subjects” into “data objects”Footnote ¹²³ and risks violating their agency, autonomy and dignity.Footnote ¹²⁴ This is a fundamental design problem that is amplified as digital solutions proliferate.

Effective digital solutions also demand strong data and cyber security.Footnote ¹²⁵ Despite investments in recent years, humanitarian organizations’ growing “cyber perimeter” increasingly exposes them to data leaks and cyber security incidents.Footnote ¹²⁶ The January 2022 data breach that affected the International Committee of the Red Cross (ICRC) illustrated the “infinite vulnerability”Footnote ¹²⁷ of humanitarian digital ecosystems to these growing threatsFootnote ¹²⁸ – and the chimeric nature of cyber security.Footnote ¹²⁹ Other incidentsFootnote ¹³⁰ have highlighted how human errors and lack of digital security awareness remain the weakest cyber security link.Footnote ¹³¹ The emerging question is whether humanitarian organizations should continue increasing their reliance on digital tools and processes when they do not have the means to effectively control them or be accountable for them,Footnote ¹³² at serious costs for the safety and dignity of the populations that these organizations are meant to protect. As recently put by a cyber security expert,

In his commentary on the humanity principle, Jean Pictet highlights that “restorative action … must be accompanied by preventive action”.Footnote ¹³⁴ Understanding that the digital transformation brings new dimensions of risks that demand more preventive efforts to avoid creating more suffering and protection needs has become urgent. In contemporary humanitarian settings, “doing no harm” means addressing the data and cyber security dimensions of human safety, security and dignity.Footnote ¹³⁵ As the gap between policy and practice continues to grow,Footnote ¹³⁶ humanitarians should pause and reflect on how to reduce it.

When “big data” turns into “bad data”

Early on in the digital transformation, technological advancement seemed like a solution to the problems described above. In the 2000s, the concept of “big data” emerged as a powerful way to gather insights from a large variety of data sources and bridge the information gaps that jeopardized public action's relevance and effectiveness.Footnote ¹⁴¹ By leveraging and combining the multitude of data generated by digital technologies, the hope was that “big data analytics” would help “find unexpected connections and correlations”, “make unusually accurate predictions”Footnote ¹⁴² and support better-informed decision-making and practices. Computerized and algorithmic management of an “overwhelming amount of information” could help to capture the data falling through the cracks of existing analogue processes.Footnote ¹⁴³ While the use of data and analytics was not new to the administration of public action, the advancement of computing capacities that came with digital innovation made everything easier and faster.Footnote ¹⁴⁴

This information “management revolution”Footnote ¹⁴⁵ was attractive for humanitarian operators struggling to make good use of the data collected through their activities, and to leverage external information to better understand the complex dynamics of their operating environments.Footnote ¹⁴⁶ The United Nations (UN) and other international organizations rapidly committed to leveraging big data and placed it at the centre of their operational and development strategies.Footnote ¹⁴⁷ This triggered a systemic shift in which data moved from the periphery to the centre of humanitarians’ agenda and practices, including vis-à-vis donors and partners, who increasingly demanded access to the data and “evidence base” supporting strategic and programmatic choices.Footnote ¹⁴⁸ Like in the private sector, data became “the new oil” and a key “value extraction” tool to support progressFootnote ¹⁴⁹ and “boost humanitarian investments”.Footnote ¹⁵⁰ If the analogy between oil and data has been criticized for its limits,Footnote ¹⁵¹ it is nevertheless helpful to understand that like oil, data is a finite resource which can similarly contaminate and damage the environment.Footnote ¹⁵² This is also true in the humanitarian environment.

The 2014 international response to the Ebola pandemic in West Africa illustrates this problem.Footnote ¹⁵³ As this example highlights, instead of supporting better-informed decisions, the multiplication of data flows “invites the problems of digital systems into the most fragile and vulnerable environments in the world”,Footnote ¹⁵⁴ outpacing the capacity of disaster responders to make sense of themFootnote ¹⁵⁵ and adding significant layers of complexity to coordination.Footnote ¹⁵⁶ Instead of supporting prevention and preparedness, big data can create an overload of not-so-relevant information that can obscure the ability of responders to see, interpret and respond to factual realities.Footnote ¹⁵⁷ The inherent limitations of data, which are often at best incomplete and at worst inaccurate, are necessarily amplified through the magnifying effect of big data analytics. This can lead to ineffective or counterproductive interpretations, turning “big data” into “bad data”Footnote ¹⁵⁸ and reinforcing existing inequalities or creating new forms of discrimination.Footnote ¹⁵⁹

It is now commonly agreed that AI systems have an intrinsic and significant bias problem.Footnote ¹⁶⁰ This well-documented issue originates from the datasets feeding algorithmic and machine learning systems, which reflect the systemic discrimination and inequalities embedded in the societal realities they aim to capture.Footnote ¹⁶¹ This is particularly true for raceFootnote ¹⁶² and genderFootnote ¹⁶³ discriminations that are so deeply entrenched in societies and so well reflected in generative AI systems.Footnote ¹⁶⁴ The problem with these algorithmic biases is that, unlike their human counterparts, they are projected at scale and are often more difficult to identify, explain and rectify.Footnote ¹⁶⁵ The internal functioning of algorithms is most often a “black box” commercial secret, carefully protected by the private companies who own these systems.Footnote ¹⁶⁶ Even the engineers and scientists who create the algorithms can have difficulties deconstructing their opacity and explaining how they transform inputs into outputs.Footnote ¹⁶⁷

These transparency and explicability issues are particularly problematic in the humanitarian domain because they limit humanitarians’ ability to explain the causal relationships between needs and responses, and to demonstrate the objectivity of the assessments they rely on.Footnote ¹⁶⁸ To demonstrate their impartiality, humanitarians must be in a position to explain the proportionality of their response to existing needs – in particular why some are addressed and others are not. When data and algorithmic biases are unidentifiable, they can hide and spread, contaminating needs assessment models, obstructing the ability to explain how proportionality was evaluated and threatening the impartiality of responses. In practice, they risk defeating humanitarians’ efforts to be more accountable for the use of the precious resources they have, and towards the populations they serve.Footnote ¹⁶⁹ The proliferation of AI systems in humanitarian action can therefore bring more, not less, opacity and problems for impartiality, accelerating “the divide between the haves and the have nots”.Footnote ¹⁷⁰

Digital divides and exclusions

The second tension that increasing digitalization introduces to humanitarian impartiality consists in the “digital divides” and exclusions resulting from the digital transformation. In 2023, an estimated 2.7 billion people, roughly a third of the global population, remained without access to digital connectivity.Footnote ¹⁷¹ These divides are unequally distributed across the globe, and are more present in Africa, the Middle East and Asia, where a large share of global humanitarian needs and operations are also concentrated.Footnote ¹⁷² Digital divides disproportionately affect people living in rural and hard-to-reach areas, who also constitute a significant share of people in need of humanitarian assistance.Footnote ¹⁷³ These quantitative divides are compounded by qualitative ones. Women and girls,Footnote ¹⁷⁴ people with disabilitiesFootnote ¹⁷⁵ and people with low levels of educationFootnote ¹⁷⁶ tend to be more excluded from digital connectivity. Many others do not have access to “meaningful” and reliable connectivity,Footnote ¹⁷⁷ expanding the nature and impact of digital exclusion to a large share of connected people who access the internet through unsafe digital devices or infrastructures.Footnote ¹⁷⁸

These compounded divides leave billions of people “who do not routinely engage in activities that big data and advanced analytics are designed to capture” on the digital periphery.Footnote ¹⁷⁹ In addition to the algorithmic errors discussed above, these “big data exclusions” result in “another type of error that can infect datasets …: the systemic omission of people who live on big data's margins, whether due to poverty, geography, or lifestyle, and whose lives are less ‘datafied’[,] … distorting datasets and, consequently, skewing the analysis” on which humanitarians increasingly depend to assess needs and prioritize their responses accordingly.Footnote ¹⁸⁰ These exclusions create a “new kind of voicelessness” and have profound impacts on already marginalized people and communities in terms of representation and inclusion, potentially jeopardizing their access to impartial humanitarian assistance.Footnote ¹⁸¹

In response, the international community has engaged in a new effort to achieve universal and meaningful connectivity by 2030.Footnote ¹⁸² The UN High Commissioner for Human Rights recently called for a new right to access the internet,Footnote ¹⁸³ and an increasing number of civil society organizations are demanding better legal recognition of the fundamental role that digital connectivity can play in enabling access to health care, education, workFootnote ¹⁸⁴ and other services essential to people's survival and well-being in humanitarian situations.Footnote ¹⁸⁵ They are, unsurprisingly, supported by tech companies,Footnote ¹⁸⁶ which are already providing “free” connectivity in developingFootnote ¹⁸⁷ or conflict-affected countries.Footnote ¹⁸⁸ Humanitarian organizations are joining the effort, developing “connectivity as aid” delivery capacities, despite the significant operational and ethical dilemmas implied.Footnote ¹⁸⁹ Yet, in parallel, others are calling for a “right to not use the internet”, observing that the digital transformation has turned connectivity “into a de facto obligation for anyone” who wants to exercise their fundamental rights.Footnote ¹⁹⁰ While many are demanding more of it, others are asking to be able to opt out from connectivityFootnote ¹⁹¹ because it “helps reproduce … inequality and external control rather than ameliorate such conditions”.Footnote ¹⁹² Acknowledging these people's perspective and respecting their choice is an increasingly relevant consideration in the difficult exercise of ensuring humanitarian action's impartiality in the digital era. In other words, digital connectivity should be a genuine choice – available for those who want it, and not required for those who do not. Humanitarian aid should be adapted to both, so that their needs, voices and perspectives are equally taken into account in the difficult proportionality assessment exercise that is required by the principle of impartiality.

Humanitarians and the geopolitical digital chessboard

Digitalization is making humanitarian neutrality and independence more difficult. It is indeed often not enough for humanitarian actors to be neutral and independent – they must also be perceived as such. In the current competitive and polarized environment, the decision to use a particular technological tool is likely to be increasingly perceived as a political choice, and one associated with those behind the tool.Footnote ²⁰² Choosing a US-based tech provider is not the same as choosing a Chinese or European one, because each provider abides by a different legal and political framework representing the preferences of its associated State. Indeed, in the increasingly competitive international environment described by Eric Schmidt at the beginning of this article, tech is about power, and therefore politics.Footnote ²⁰³ States and private actors are racing to control digital technologies and the supply chains behind them – from the rare-earth materials from which microchips are madeFootnote ²⁰⁴ to the skills and machines required to build those chips, and all the way to the infrastructures and data that are needed to make them functionFootnote ²⁰⁵ – and are building coalitions to increase their commercial and political influence at the global level.Footnote ²⁰⁶ While this competition is not new, the stakes are increasing, as is the impact on the humanitarian sector.Footnote ²⁰⁷

When donors, host States or partners are promoting or demanding the use of specific technologies or brands in the context of humanitarian action, this reduces the choice that humanitarian organizations can make to select the tools that best fit their needs and operational constraints, thus de facto challenging their operational autonomy and independence. When those technologies or brands also happen to be used for military or security purposes by parties to conflict or entities associated with them, there is a risk that such a choice will be perceived as a political one, thereby impacting the perception of humanitarian organizations’ independence from those parties and entities and jeopardizing the perception of their neutrality. As the current debates on sanctions against certain Chinese companies and technologies demonstrate,Footnote ²⁰⁸ States are using sanctions and other restrictive measures to influence who can use what technologies. Just as sanctions in other domains contain exceptions to exempt humanitarian actors from their scopeFootnote ²⁰⁹ so as to preserve the perception of their independence and neutrality from political decisions, it seems increasingly relevant to also preserve such actors from undue political interference or cooptation through the choice of technology used for humanitarian purposes. Indeed, humanitarians’ growing reliance on digital technologies is an interesting vector for expanding the geo-strategic digital battle into new territories, and humanitarian organizations risk becoming de facto tactical pawns on the international chessboard for digital hegemony.Footnote ²¹⁰ Understanding the digital dimensions of neutrality and independence can help them find a way out of a game they are not meant to play. This starts by understanding the digital transformation's political economyFootnote ²¹¹ and discerning the roles that its key actors play in order to better delineate the parameters through which humanitarians should relate to those actors in their efforts to best protect the perception of their neutrality and independence.

First, the origin of most truly innovative digital technologies over recent decades can be traced back to the research and development (R&D) activities of the main actors of conflicts: States’ armed and security forces. The internetFootnote ²¹² was created in the laboratories of the US Department of Defense's Advanced Research Projects Agency (ARPA).Footnote ²¹³ Drones were born on the battlefield to support air warfare capabilities.Footnote ²¹⁴ The machine learning tools that power commercial AI applications are often inspired by innovations geared for military purposes, and the “Turing test” that defines their level of “intelligence” was invented in a military context.Footnote ²¹⁵ Today, facial recognition and biometrics are mostly used for security purposes, from identifying potential “terrorists” to managing border security and migration flows.Footnote ²¹⁶ In short, if many digital technologies are “dual-use” (i.e., both military and civilian) in nature, it is important to remember that they were often initially created and designed for military and security purposes, and that they are increasingly central to States’ agendas and strategies in these domains. When humanitarian actors choose to use the same technologies or tools as political and military actors, they should take into account that affected people and parties to conflict may perceive such a decision as an affiliation by association with those actors – even though in practice, they often have no choice considering the ubiquity of those technologies or brands and the absence of any truly adequate alternative choices.

Over the years – and despite significant increases in States’ defense budgets across the worldFootnote ²¹⁷ – the R&D capacities of States have often been outpaced by those of tech companies, which have thus become critical strategic partners in the innovation race. These relationships have grown closer – and more blurred. In China, Russia, the United States, Europe and elsewhere, States have always been important investors for tech companies, but the digital transformation has brought their symbiotic relationship to a new level.Footnote ²¹⁸ Tech companies’ investment and governance structures are evolving accordingly, causing some to describe the traditional distinction between the public and private sectors in the tech domain as a “myth”.Footnote ²¹⁹ The “revolving door” between tech companies and governments, which are alternatively managed or advised by the same individuals, illustrates the growing difficulty – including for humanitarian actors in search of neutrality and independence – of distinguishing between political and commercial actors and delineating their respective agendas.Footnote ²²⁰

This blurring of the lines is also reflected in tech companies’ growing influence in conflict and humanitarian settings.Footnote ²²¹ From supplying connectivity to the deployment of digital means of warfare,Footnote ²²² to providing technical advice and support for cyber defence,Footnote ²²³ to complying with government requests for data or “takedowns” of information content on their platforms,Footnote ²²⁴ tech companies increasingly find themselves in the middle of the battlefield.Footnote ²²⁵ Sometimes this is largely involuntary and due to the dependencies attached to commercial partnerships with States, or to ownership of digital infrastructures.Footnote ²²⁶ In other situations this alignment is triggered by a strategic decision to support parties to a conflict,Footnote ²²⁷ in line with the tech company's interests or values.Footnote ²²⁸ In recent years, tech companies have proactively invested in their capacity to influence the global political agenda, including its security dimensions.Footnote ²²⁹ Tech companies’ founders and CEOs – and a growing number of lobbyists working on their behalf in places of powerFootnote ²³⁰ – are leveraging the space left by the erosion of traditional multilateralism in order to advance their own agendas,Footnote ²³¹ and are becoming increasingly political actors from which humanitarian organizations should maintain as much independence as possible.

Tech companies’ objectives are multiple and diverse, sometimes overlapping with and sometimes diverging from States’ – and humanitarian – objectives. This complexifies the traditional independence principle. In some cases, tech companies’ objectives (which have included moving people to MarsFootnote ²³² or acquiring guns and a safe refuge in case AI drives humanity to its extinctionFootnote ²³³) can overlap, or create tensions, with humanitarian objectives. For instance, the “effective altruism”Footnote ²³⁴ or “long-termism”Footnote ²³⁵ allegedly driving the use of digital technologies for social goodFootnote ²³⁶ seem generally aligned with humanitarian ambitions, but in practice, the methodologies employed in such projects raise significant questions about their short-term impact and the negative consequences they are already having.Footnote ²³⁷ Some of these projects, involving the collection of iris scans from populations with limited income in exchange for digital identities and currencies,Footnote ²³⁸ have raised data protection alarmsFootnote ²³⁹ and concerns regarding data extractivism.Footnote ²⁴⁰ The humanitarian marketing veil used to promote these private sector initiatives feeds “hybridization” concernsFootnote ²⁴¹ because such projects can create a dangerous confusion with purely humanitarian endeavours, in particular when they are carried out through partnerships with the tech companies behind the systems involved.Footnote ²⁴²

The evolution of tech companies’ political posture and ambitions changes their identity and perception, and this impacts their relationships with humanitarian organizations, with potentially negative consequences for the perception of the latter's neutrality and independence from associated political and strategic objectives. Tech companies’ dominance over the digital transformation amplifies the asymmetries of power that characterize their partnerships. It also increases the risks of “aidwashing”, through practices that “involve the use of corporate social responsibility initiatives and … partnership with aid actors to burnish surveillance firms’ reputations and distract the public from corporate misbehaviour, ethical misdeeds, and dubious data practices”.Footnote ²⁴³ The problem of “dual loyalty” that can deflect humanitarians’ “primary loyalty to … those affected by crises” towards the third-party digital service providers on which they rely has “real implications for the rights and needs of affected people”.Footnote ²⁴⁴ While the trade-offs attached to these partnerships can be positive in terms of efficiency and scale, they can also negatively impact the perception of humanitarians’ independence and the attendant trust of the populations they serve.Footnote ²⁴⁵

While humanitarian organizations battle to reconcile their history with renewed and legitimate questions about colonialism,Footnote ²⁴⁶ their contribution to the expansion of the digital transformation of fragile countries is fuelling concerns about their independence.Footnote ²⁴⁷ The growing debates around the role of humanitarians in “techno-colonialism”Footnote ²⁴⁸ and “digital extractivism”Footnote ²⁴⁹ highlight the concerns that the historical intersections between colonialism and humanitarianism are repeating along the same routes and power asymmetries.Footnote ²⁵⁰ Instead of natural resources and workforce, “techno-colonialism” aims at extracting data from the “digital bodies”Footnote ²⁵¹ of people and communities of the global South in order to fuel the “fiefdoms”Footnote ²⁵² and data-hungry surveillance business models of multinational tech companies. While these concerns are debatable, they call for serious examination by humanitarian organizations that have placed “localization” high on their agendasFootnote ²⁵³ but are technologically moving away from it.Footnote ²⁵⁴

Anecdotal examples of how digitalization is impacting the perception of humanitarian actors’ neutrality and independence are multiplying. The example of the UN World Food Program (WFP) partnership with the data analytics firm PalantirFootnote ²⁵⁵ – a key partner of many security agencies across the worldFootnote ²⁵⁶ – illustrated the growing concerns around “surveillance humanitarianism”.Footnote ²⁵⁷ There is indeed a “significant but little understood” risk that these partnerships are used by the military and surveillance tech industry to gain access to strategic information, new markets and data streams,Footnote ²⁵⁸ notably through national security-based legislation enabling possible “backdoors” or data access requests.Footnote ²⁵⁹ In conflict settings, that risk has already been identified by authorities who allegedly stopped WFP's assistance programmes due to concerns about the further use of their biometrics registration data.Footnote ²⁶⁰

Humanitarians are familiar with accusations of spying and partiality, which are common in polarized conflict settings. Regrettably, such accusations can become more difficult to reject when humanitarian actors rely on the same tech suppliers as conflict parties for data storage or connectivity.Footnote ²⁶¹ This suspicion grows when they partner with tech companies engaged in military activities.Footnote ²⁶² When tech partners proactively become involved in conflict-related issues, humanitarian organizations become de facto hostages of those relationships, impacting their independence and the perception of it.Footnote ²⁶³

Digital dependencies and the “splinternet”

The growing dependenciesFootnote ²⁶⁴ that come with the digitalization of humanitarian action also trigger certain operational challenges. First, dependency on proprietary hardware and software to run operations creates a “vendor lock-in effect” that decreases the ability to choose alternative tools or providers – for instance due to perception concerns – without substantial switching costs.Footnote ²⁶⁵ Digitalization comes with a “ratchet effect”Footnote ²⁶⁶ that makes it difficult to stop using digital tools once they are integrated into operational structures; for example, cloud data storage is now often required to support digital services platforms at scale. This is particularly true for digital solutions that have been put in place for exceptional circumstances (such as temporary lack of access due to insecurity or pandemics): they tend to stay when those circumstances disappear, often because the investments behind them have significant amortization costs. In practice, such solutions have also eroded the operational resilience of humanitarian organizations, which have often disinvested in analogue alternatives – everyone uses a smartphone to coordinate field activities, but many do not know how to use a VHF radio, which may be life-saving when connectivity is down. In a world where internet shutdowns and connectivity denials are on the rise,Footnote ²⁶⁷ humanitarians should not over-invest in digital technologies at the cost of the ability to operate in low- or no-connectivity settings.Footnote ²⁶⁸

Another threat for operational independence is the fragmentation affecting the digital transformation backbone.Footnote ²⁶⁹ Once celebrated as a globalized level playing field, the internet has become a divided and contested space where States battle to assert their “digital sovereignty”.Footnote ²⁷⁰ While States struggle to align at the international level, they advance their respective strategies through massive investment campaigns (such as the Chinese Digital Silk Road initiative),Footnote ²⁷¹ protectionist measures (such as US import/export controls and sanctions)Footnote ²⁷² or regulatory action (such as the extraterritorial “Brussels effect” of the European Union GDPR or Digital Services Act).Footnote ²⁷³

The internet is turning into what some refer to as a “splinternet”Footnote ²⁷⁴ – i.e., a fragmenting digital and cyber space. This is making the internet increasingly difficult to navigate for humanitarians. One can imagine the practical challenges of running digital solutions in Africa, where the digital infrastructures are more likely to be Chinese, with technological tools often provided by US tech companies,Footnote ²⁷⁵ and data protection or sanction requirements from European donors. The absence of harmonized regulatory standards and interoperable technologies could become a real obstacle to the continuity and effectiveness of digital solutions.Footnote ²⁷⁶ To preserve the perception of their independence, it has become essential that humanitarian organizations anticipate these developments and prepare accordingly. An increasingly digital future requires them to upscale their ability to do so.