Module 4, Nov 18, Seminar and lab. The Digital Story is due this week.

Design justice and ICT4D

Background

The affirmative case starts with a database. DHIS2, an open-source health information system built by the University of Oslo and national health ministries, grew from a 1994 action-research project in post-apartheid South Africa into software running in more than 70 countries covering about 2.4 billion people; a 2025 review comparing DHIS2 with FAIR data approaches in Uganda found near-national coverage with around 98 percent data completeness, and the Oslo group’s own analysis credits the survival to distributed local adaptation instead of any launch-day design. The most famous ICT4D result also supports the affirmative case. A 2016 study in Science found M-Pesa, Kenya’s mobile-money system, lifted about 194,000 households out of poverty.

Survival at that scale is the exception in its own field. A scoping review of 738 documented digital-health interventions across sub-Saharan Africa found only about half could be classified as established, most oriented toward data extraction rather than better service delivery, and 92 percent demanding time from already-stretched health workers. Researchers who study DHIS2 as a digital public good also document its scaling paradoxes, with donor priorities crowding out local user voice, and propose subsidiarity, decisions made at the most local effective level, as the governance corrective.

The critical case answers on two fronts. A 2019 peer-reviewed rebuttal argues the M-Pesa finding rests on method problems and that the product sits inside a fintech economy driving household over-indebtedness. And participation itself has a measured gap between principle and practice: Sloane and colleagues call it participation-washing when consultation masquerades as power transfer, and a field evaluation of participatory infrastructure design in Makassar, Indonesia and Suva, Fiji found high overall attendance, 89 and 75 percent, with sharp gaps underneath: disabled residents in Suva were 92 percent less likely to attend the workshops, and religious minorities and unmarried residents had less influence over final decisions even when they attended. Good intentions and open processes exempt nothing.

Key ideas

Design justice

Sasha Costanza-Chock’s argument is that every interface assumes a default user, and that unmarked user tends to be the most privileged one. Designed systems distribute affordances and disaffordances unequally across what she calls the matrix of domination, usually without anyone intending it; her cases include Facebook’s algorithmic suppression of non-Western names and virtual assistants failing to recognize disclosures of sexual assault. Design justice responds by asking who participates in design, who is accountable, and whose knowledge counts, and the Design Justice Network’s ten principles, published in 2018 as a living document, recast designers as facilitators and treat everyone as an expert based on lived experience. Her chapter title ends in a question mark (“Hard-Coding Liberation?”) because the process around the object and the institution it lands in do most of the deciding.

Social work already had a version of this claim, developed decades earlier for direct practice with clients. Lorraine Gutiérrez’s empowerment framework, built from her work with women of color, moves practice through four steps: increasing a person’s sense of self-efficacy, developing group consciousness, reducing self-blame, and assuming personal responsibility for change. Gutiérrez taught at the University of Michigan School of Social Work and is now professor emerita; her framework is the discipline’s own root for the participatory claim Costanza-Chock makes from the design side.

The participatory tradition behind all of this keeps auditing itself. Bødker and Kyng argue that participatory design has confined itself to small, comfortable projects while Facebook, Uber, and Mechanical Turk reorganized everyday life without it, and they call for researcher-activists who stay with a deployment past the prototype stage. Participatory action research with an Australian regional community reached a related verdict on packaged design-thinking workshops: they work as a creative warm-up and rarely produce systemic change. The authors’ name for that failure is innovation theatre.

The question is arriving in social work practice directly. A 2025 participatory design study with social-service practitioners ran co-design workshops with 27 participants and contextual inquiry with 24 practitioners on generative AI in social services, and found practitioners want the tools for documentation, assessment, and worker supervision while naming overreliance, algorithmic bias, skill atrophy, and client safety as the risks.

Design from the margins

Afsaneh Rigot’s method flips the usual order: build for the most decentered, highest-risk user from the start, as the primary case instead of an edge-case retrofit. Her cases are people whose safety depends on software behaving well under hostile conditions, among them LGBTQ+ people under state surveillance, activists, and sex workers, and her load-bearing claim is generalizability: what is created for the decentered cases works for the broader usership too. Systems designed this way turn out safer and more private for everyone else.

Networked movements and digital organizing

Design justice asks whose narratives carry authority, and networked media is one place that authority gets contested. Jackson, Bailey, and Foucault Welles show how marginalized groups use hashtags such as #MeToo, #SayHerName, and #Ferguson to build counterpublics and challenge how mainstream media frames them, and Freelon, McIlwain, and Clark, analyzing 40.8 million tweets, found Black Lives Matter activists set the narrative agenda largely independent of the mainstream press. The limits are documented too. Zeynep Tufekci argues that networked movements scale quickly without building the organizational capacity that older movements developed slowly, so they can peak and then stall, and studies of token participation find that public gestures of support can substitute for action rather than lead to it, as when 99.76 percent of a 1.2-million-member Facebook campaign never donated.

Social work has taken this up as macro practice. Brady, Young, and McLeod studied a largely online worker-rights campaign and concluded that social media is a useful organizing tool but not a replacement for relational organizing, and a 2021 special issue of the Journal of Community Practice argues that digital macro practice has become a core social work skill rather than an optional one.

Toyama’s amplification thesis

Kentaro Toyama’s amplification thesis is the counterweight to both design traditions: technology multiplies existing human and institutional intent and capacity. A capable, caring institution paired with a good tool gets better outcomes, and an extractive or broken institution paired with the same tool gets faster extraction. His evidence is a null result at national scale: across the decades of continuous digital innovation since the 1970s, US poverty has not declined and inequality has increased dramatically. Toyama teaches at the University of Michigan School of Information.

The thesis has a canonical teaching case. Namank Shah’s post-mortem of One Laptop Per Child argues the initiative failed by importing a Western machine onto communities without adapting to local needs or culture: Uruguay’s laptops shipped with music software that could not handle local musical forms, distribution targets were never realistic, and the design theory that children would teach themselves sidelined the teachers who decide whether classroom technology gets used at all.

Commons and capabilities

Benkler names the production model behind Wikipedia, OpenStreetMap, NVDA, and DHIS2: commons-based peer production, strangers building a shared resource nobody owns. The idea has a 1973 ancestor in Ivan Illich’s Tools for Conviviality, which distinguished convivial tools, extending individual and community capability, from manipulative tools, which concentrate control in institutions. A working current example is Missing Maps, founded by the American Red Cross, British Red Cross, Humanitarian OpenStreetMap Team, and Médecins Sans Frontières and now spanning 17 member organizations: remote volunteers trace satellite imagery into OpenStreetMap, community volunteers add local ground detail, and humanitarian responders use the resulting maps.

The normative ground under “good” in ICT4D comes from Amartya Sen’s capability approach, development as the expansion of people’s real freedoms. Kleine’s Choice Framework turned it into a usable evaluation method, worked out through a case study of rural telecentres in Chile, that asks whether an intervention expanded the community’s actual choices.

The current institutional form is the digital public good, with India’s UPI payment infrastructure as the flagship case and a live debate over whether the “public” framing masks continued private capture; Purandare argues that citizens become data points unless internet access itself is treated as a right requiring state investment. The discourse has its own access problem: at the 2024 Civil 20 summit, a breakout session on digital public infrastructure drew two participants until it was renamed “digital government”.

Digital identification systems

The largest live test of these arguments is biometric ID. India Stack’s e-KYC layer verifies identity by fingerprint or iris before a service is delivered, including at the ration shops that distribute subsidized food. An India Development Review investigation documents what verification failure means there: the 2020 State of Aadhaar report found more than 30 percent of people who hit an authentication failure at a ration outlet received no food at all, and by March 2025 only 75.69 percent of ration-card holders nationally, and 6.6 percent in Delhi, had completed e-KYC. The failure points it names are ordinary ones: worn fingerprints among manual laborers, unreliable rural connectivity, Aadhaar-linking errors, and geographic restrictions that strand migrant workers.

Kenya’s version reached a courtroom. The High Court halted the rollout of Huduma Namba, a national biometric ID system, until a proper legal framework existed, and declared its DNA and GPS data collection unconstitutional. The court declined to find that the system discriminated against the Nubian and Somali minority communities who already struggle to obtain the documents it would require, so the users the margins framework centers are the ones the ruling left unprotected.

The M-Pesa case

In 2016, Tavneet Suri and William Jack concluded in Science that access to M-Pesa had lifted roughly 194,000 Kenyan households, about 2 percent of the country’s households, out of extreme poverty, with the strongest effects for female-headed households. The mechanism they proposed was financial resilience: mobile money made saving and transfers easier, and women in particular moved out of subsistence agriculture into business. The scale of the phenomenon is visible in Our World in Data’s mobile-money series, built from the GSM Association’s global dataset, which tracks active accounts across most world regions from 2010 to 2024 and shows negligible adoption in North America.

In 2019, Bateman, Duvendack, and Loubere published a peer-reviewed rebuttal arguing the finding rests on what they call a worrying number of omissions, errors, and inconsistencies, and that M-Pesa belongs to a wider fintech economy pushing Kenyan households toward over-indebtedness.

Both readings are peer reviewed and they reach opposite conclusions, which makes the pair a working record of how a celebrated tech-for-good result gets re-examined. In class, each group names the single piece of evidence that would move it from one reading to the other, and the debrief compares those answers.

Before class

In class

Seminar

Vote, then design justice against HCD and design thinking, with the Design Justice Network Principles as a handout. Toyama’s amplification thesis comes in as the counterweight, and the two ICT4D exhibits follow, the M-Pesa contest, with Our World in Data’s mobile-money adoption curves, and the DHIS2 case. Revote before the lab. A guest session on participatory design and health-equity HCI is planned (TBC); speaker and pre-reading confirmed on Canvas.

Lab: design from the margins, applied to your partner

Each group identifies the most decentered user its partner serves: the person the current system fits worst, whether because of language, device, disability, schedule, or documentation status. You name that user concretely, walk one existing decision (a form, an eligibility rule, a communication channel) through their situation, and redesign that one decision for them first, following Rigot’s method. The redesign and its rationale go into your pitch’s accessibility and ethics sections.