Beyond Digital Financial Services: Exploring Mobile Money Agents in Tanzania as General ICT Intermediaries

Tanzanian mobile money and telecom agents (called wakala(s) in Swahili) have played a crucial role in expanding digital financial services (DFS) to rural areas. However, wakalas are losing their ability to financially sustain themselves providing intermediation services that their communities require. This work explores the potential for the wakala network to extend intermediation services to emerging information and communication technologies (ICTs) beyond the scope of commercial DFS by uncovering the social and institutional factors that currently shape wakala practices. First, we investigate how two different models of intermediation from information and communication technologies for development literature can inform broader strategies for intermediation through human infrastructures. We then complement this analysis with an on-the-ground quantitative survey and focus groups with community members and wakalas in Kagera, Tanzania. Our focus groups reveal that community members face challenges with new ICTs that require sustained intermediation and that wakalas encounter mounting financial instability and are thus receptive to intermediating for other ICTs. Finally, we present three factors that influence the broadening of the wakalas’ role of general ICT intermediaries: (1) aligning incentives and addressing the limits of pro bono actions, (2) providing appropriate training and a suitable support infrastructure, and (3) fostering trust-building and reciprocity.


INTRODUCTION
In the Global South, information and communication technologies (ICTs) promote inclusive and equitable access to information and services where existing market structures leave gaps.However, research shows that ICT performance over time falls short of expectations [2] due to varied 3:2 A. Raghunath et al.
factors: user and provider lack of access to technology, infrastructural barriers, low digital and language literacy, cultural factors, costs, and lack of training and sustained support to help users learn and navigate systems.It is clear from the literature and many deployments that selecting, training, and supporting intermediaries plays a critical role in introducing and sustaining ICTs in contexts where familiarity with technology is low [60,70].
The proliferation of mobile money services in Sub-Saharan Africa is a notable exception, where ICTs have made substantial progress in giving un-and under-banked populations access to digital financial services (DFSs) through their mobile phones.This success has been credited to mobile money agents (hereafter referred to by the local Swahili term, wakalas) in rural and hard-to-reach areas, instrumental contributors to increasing financial inclusion in many East African markets.Wakalas-local retailers who double as "low-cost" intermediaries to banking services-help customers access saving and deposit products without visiting brick-and-mortar banks.
The wakala human infrastructure network converts physical cash to digital value and performs crucial tasks, such as on-boarding, supporting and educating millions of customers on a daily basis.However, data show that increases in digital payments effectively minimizes the role of human agents.Between 2012 and 2021, the amount cashed out from every dollar that was deposited into mobile money accounts dropped from 88 cents to 67 cents, and the mean monthly volume of cash withdrawals and deposits made by agents in sub-Saharan Africa declined from $11,700 in 2016 to $9,900 in 2021 [4,16].This declining activity means fewer opportunities for wakalas to earn commissions on routine DFS transactions.
Amidst increased competition and lowering returns, wakalas are therefore relying on other businesses to compensate for the loss in mobile money revenue [22].Their cultivation of skills and trust relationships with community members over time may make them appropriate intermediaries for new ICTs that provide a viable way to supplement their income.
The preceding problems and concerns motivate the following research questions we explore here: RQ1: Given the changes in the DFS ecosystem, what are current wakala intermediation practices in their communities?RQ2: What key factors influence wakalas' interest and ability to intermediate for different ICTs that remain underutilized?
To answer these questions, we adopt a mixed-methods approach that uses a quantitative survey and focus groups (FGs) with community members and wakalas across five villages in Kagera, Tanzania.Our survey reveals that wakalas offer a popular form of structural support for rural Tanzanian phone users when accessing DFSs and other technologies, more generally.Our FGs explore the intermediation offered by wakalas from both their perspective and that of community members they support; among our findings, we highlight the intricacies in defining who needs sustained support with both new and old ICTs, and challenges in how support and trust is enacted within different incentive schemes.Further, we examine how the Tanzanian government could optimally leverage the wakala network to speed proliferation of ICTs and establish a well-supported community that feels agency to use new technologies.
Our findings demonstrate the opportunities and obstacles that accompany the goal of expanding wakala infrastructure beyond the current DFS domain, making the following contributions to the information and communication technologies for development (ICTD) community: perceptions about phone capability to deliver said information to non-technical or illiterate users, and the ability of rural communities to trustfully use ICTs in East Africa [2,73].

RELATED WORK
We now review the literature from ICTD and human-computer interaction (HCI) on the role of intermediation in increasing uptake of digital technologies (Section 3.1) and DFSs (Section 3.2).We then contextualize our work in the existing literature on human infrastructures and their relationship to maintaining DFS systems (Section 3.3) and examine two different models of intermediation, seeking insights about successful intermediation implementations (Section 6).

Intermediation
ICTD and HCI literature have examined the ways in which intermediaries shape the design, use, and impact of technology, including ICTs in the Global South.Parikh and Ghosh [51], Sambasivan et al. [58], and many others shed light on intermediation as a fundamental component of technology use and access for a vast number of people in low and middle income countries (LMICs).Bailur and Masiero [5]'s study maintained that intermediaries should "attempt equitable participation, be trusted by the community, and preferably be 'local' and technically proficient." This definition of intermediary guides our study.
Chaudhuri [11] discusses how "digital technologies are expected to create a direct link between users and information." However, "illiteracy, low levels of education, gender, class and caste inequalities, and other social endowment gaps become significant barriers, necessitating intermediaries." Similarly, Oreglia and Srinivasan [50] debunk "free" information flows enabled by ICTs, shedding light on the actors and practices that enable financial transactions by agricultural intermediaries.Parikh and Ghosh [51] put forth a taxonomy of intermediated information tasks, ranging from cooperative to dominated, intermediated, or indirect.While intermediation typically relies on a physical intermediary, Meissner [37] formulated the idea of intermediating through the internet and text messaging to support areas low in human capital.These forms of intermediation can be further specified to suit low-literate individuals, minority genders, specific cadres of workers, and so on, and different types of intermediaries.
Several examples show intermediaries to be successful agents in broadening access to ICTs.For example, Kumar and Anderson [32] recognize youth as helpful intermediary to ASHAs (community health workers) using digital health tools in India; Madaio et al. [34] find parents intermediating educational technology on behalf of children with the goal of improving their literacy; Ogbonnaya-Ogburu et al. [49] discuss the intermediated learning of technology by formerly incarcerated individuals; and Wahid et al. [71] describe religious groups as intermediaries in Indonesia.Some technologies are built specifically with intermediation as the predominant use case, potentially even conferring the "primary user" label to the intermediary [51].For example, Mehra et al. [36] built an intermediated smartphone app to present loan information to auto rickshaw drivers; Rahman et al. [53] leveraged small pharmacy owners as intermediaries on an Android app to close the information gap about healthcare resources for impoverished Bangladeshis; and Seth [60] uses "community mobilizers" to onboard new users, create content, moderate, and build community on their social platform, and so on.
Despite the importance of intermediaries in ICT delivery and uptake, government systems and literature have explored moving beyond agents to create sustainable ICTs.Rohanifar et al. [56] find that "dishonest and unfair intermediaries" were "a hindering factor for the slow acceptance of financial systems in the context of Bangladesh, " and the Aadhar framework in India was standardized to minimize the role of intermediaries and thus limit corruption in the last mile delivery of government services.Our work adds nuance by uncovering how intermediary incentives could change given different market structures, how technical systems could be strengthened to afford intermediaries and community members a more robust ICT ecosystem, and how Tanzanian intermediaries' skills and trust relationships could be scaled and translated to promote adoption of ICTs beyond DFS.

Digital Financial Services and Intermediation
Since 95% of the world's population is covered by a mobile broadband network, branch-less banking or agent banking have become important tools for delivering financial services to the poor [6,10,12,52].Though research on agent banking has recognized the importance of intermediaries in supporting financial inclusion goals [17,26,31], it has also consistently highlighted the precariousness of this model [8,26,31,69]: the model's viability and long-term sustainability is limited by market size. 1herefore, since most local agent models are implemented in a franchise-like model [69], with agents compensated based on the targets they meet rather than a fixed salary, they compete against each other for limited numbers of customers, which strains their incomes.This lets the principals (financial service purveyors) increase labor (agents) without assuming the fixed cost associated with this expansion.This dynamic, in addition to other factors that influence market size (urban/rural, transport cost, scope), contributes to the low profitability observed in local agent revenue [17,31].Our work finds similar evidence of this dynamic among wakalas in Tanzania and explores how ICTD researchers and policy-makers can sustainably leverage the human infrastructure they provide to mutually benefit wakalas and consumers.

Human and Technological Infrastructures of Digital Financial Services
Researchers in ICTD have recognized the importance of analytically framing our understanding of infrastructures [1,14,25,57,66].From identifying the role of Anganwadi workers in India [57] to trust in biometric infrastructures [35], this view promotes the need for the ICTD researcher to prioritize both the technological and human systems that support the artifacts we observe.Infrastructures depend on multiple technical and social structures, embody standards, and invisibly support tasks until they can be expressly examined [28,67,68].
A key concept in infrastructure studies is infrastructural inversion, or the activity of foregrounding infrastructure to demonstrate its often hidden relational nature with other structures [64].This inversion highlights the contingent and continuous work of numerous human actors in maintaining an infrastructure, which Reference [33] refer to as human infrastructure and stress its importance in explaining how technologies are maintained.Particularly relevant to our research is the view of mobile money and other DFSs as infrastructures [21,29].In a study of the expansion of DFS aimed at extending access to loan products, Reference [21] describes the complex and seamless interaction of the mobile money agent infrastructure and the loan repayment infrastructure, highlighting the human work that makes the provision of loan services possible; the authors view mobile agents as indispensable to the effective function of loan management DFSs [21].Our work also uses human infrastructures as a lens through which to explore the role of wakalas in Tanzania.

Two Models of Intermediation: DHIS2 and Mobile Vaani
Our second research question asks how to leverage wakalas as intermediaries for technology interventions beyond digital financial services and the factors that influence this broader intermediation role.Here, we introduce two intermediation models, the District Health Information System version 2 (DHIS2) and Mobile Vaani, to demonstrate the highly complex interaction between a technology's market, the goals of its creators, and the social and financial incentives that influence intermediary actions.

The DHIS2 Model.
"The driver of DHIS success is the DHIS and Health Information Systems Program (HISP) community and network, which is built up around projects and activities carried out by university groups, NGOs, health authorities, and other actors, networks of action" [9].DHIS2, launched in 2011, is a prominent open-source Health Management Information System typically used in LMICs as national health information systems for data management, health program monitoring, logistics management, and disease surveillance and reporting [7].It has roots in action research by HISP, a collaborative program involving the University of Cape Town, the University of Western Cape and a Norwegian researcher, that aimed to address the information management challenges of fragmented healthcare system in post-apartheid South Africa; it is currently administered out of the HISP Center in Oslo and supported by numerous large international aid organizations.
At the heart of DHIS2 growth is its supporting community, which includes the HISP global network of users, developers, funders, and researchers.As the project grew, it sought to "engage students from both the health and informatics disciplines in a collaborative way in the development of health information systems in their own countries" [9].Today, DHIS2 makes continual and concerted efforts to build its global community of users and experts.For example, the DHIS2 Academy, which to date has graduated close to 6,000 participants, builds capacity through a combination of online courses and in-person training, which promise to create a ready-to-go, directly compensated (non-volunteer) DHIS2 expert consultant who can leverage the model to support any organization.
Though DHIS2 is an open-source application, "forks" of the code-base were not recommended due to fears of multiple separate and incompatible system versions.All application development on the platform is conducted only by in-house developers, third-party developers, or developers in its HISP network with existing expertise in DHIS2.This means that the work of intermediating DHIS2 requires significant training and expertise not only in the DHIS2 data model and APIs but also in its various orthogonal features, e.g., routine health information and entity tracking modules.
Today, DHIS2 makes continual and concerted efforts to build its global community of users and experts.For example, the DHIS2 Academy builds capacity through a combination of online courses and in-person training, which promise to create a ready-to-go, directly compensated (nonvolunteer) DHIS2 expert consultant who can leverage the model to support any organization.

The Mobile Vaani Model.
Mobile Vaani is a community media platform based on interactive voice response (IVR) that was created by the Indian social enterprise Gram Vaani (GV) with the goal of "empowering poor and marginalized community to create their own local media" [39].It was then popularized in different regions within the state by its volunteers or community mobilizers.After receiving additional funding in 2013, GV began investing in more robust training and recruitment of volunteers to strengthen the role of community mobilizers.
In 2015, GV began to redesign the structure of Mobile Vaani, moving away from an uncoordinated network of 120 volunteers by introducing financial incentives.The first intermediation structure, a capped reimbursement and reward scheme, was intended to compensate volunteers for out-of-pocket expenses incurred during their work and reward them for their mobilization activities (i.e., the number of users they recruited to the platform).Additionally, GV launched a referral system to track new user acquisition instead of relying purely on self-reported data.However, this structure faced significant problems due to misreporting, an abuse observed in Mobile Vaani's recall metrics, 2 which dropped from 25% to 5% [39]."The referral system was therefore discontinued, but it led to anger and resentment of honest volunteers who were making legitimate referrals and (importantly) had begun to rely on the additional income for their livelihoods" [39].
After a failed attempt to design a financial incentive mechanism for volunteers, the GV team realized that more support was need in human resource management than in developing the actual technology.Eventually, GV changed the organizational structure of Mobile Vaani from a state-level platform to a smaller federated platform managed by a volunteer club at the district level.Each club included a group of volunteers in a district that controlled a unique phone number and IVR channel to carry its local content and connect it to a wider statewide channel [39,61].The volunteer club had a four-tier structure, with community members (users), club volunteers, a club coordinator, and a community manager.Moitra et al. [39] notes that "...it was clear to the GV team that even a technology-driven pathway to development required a significant investment in human resources to supplement the technology, but managing a large team of volunteers, dispersed in remote pockets and motivated by different incentives, was a far more complex challenge than developing the technology itself!Volunteer selection criteria, internal accountability through measurable indicators, and group dynamics resilient to social conflicts were all important aspects that needed fixing." GV designed a new financial incentive structure, as well: a group incentive tied to volunteer club performance that was shared among club volunteers; an individual incentive based on each volunteer's contributions; and a self-reported individual incentive based on the number of community workshops that a volunteer conducted.Further, the volunteer selection process became more rigorous, requiring at least three months of effort in supporting Mobile Vaani before a user could graduate into the role of Club Volunteer.The authors of Reference [61] believe that this combination of group and individual incentives has allowed GV to manage the trade-off between social and monetary incentives and ensured that altruism was not deprioritized by financially incentivized tasks.

3.4.3
Takeaways.These models elicit the following observations about implications for our research.First, intermediation tends to be a highly specialized activity attuned to community needs, a technology's complexity, and the intermediary's "elasticity" (i.e., responsiveness to both financial compensation and their expected workload).
Second, successfully intermediating a technology at scale, especially if the technology's users are geographically dispersed, requires significant investment in human resources management.Third, even with open-source software and open specifications, no two technologies are alike: two forks of technologies created by the same organization can lead to significant development and management overhead that requires more intermediation to coordinate.In such cases, tradeoffs must be made between complex, feature-rich technologies that require significant expertise to intermediate and simple, highly standardized applications with a potentially lower technical barrier to entry.

METHODS
We take a mixed-methods approach that uses a quantitative survey and FGs and implemented our research in two phases: Phase I (June 2022) and Phase II (October-November 2022).Phase I encompassed 47 in-person, enumerator-led pilot quantitative surveys of households (HHs) in two randomly sampled villages in the Bukoba Rural district: Ibwera and Nyakibimbili.The survey included questions on general phone use, knowledge of USSD services, and utilization and trust of a local wakala.HHs were sampled systematically with the help of local enumerators, who walked through the village, visited one HH, and skipped six others before arriving at the next HH to screen, a sampling strategy we believe to be uncorrelated with underlying structural characteristics of the villages.When arriving at a HH, the enumerator screened the head of the household by asking their age and if the HH had a working cell phone.
In Phase II, we employed local enumerators to run 12 FGs in Swahili, encouraging them to amplify the voices of women and the elderly, and offered compensation and beverages to all participants.In total, enumerators held FGs with 70 people: 5 FGs that included 30 community members and 7 FGs that included 40 wakalas from five villages.The villages of Izigo, Bunazi, Rwagati, Kanazi, and Ngandi were cluster sampled from three districts (Bukoba Rural, Muleba, and Missenyi).For wakala sampling within each village, we employed a local guide who introduced us to both mobile money and telecom wakalas, making sure to select those from the center or main road as well as from internal areas.The HH sampling strategy is the same as we used for Phase I.
Our FGs with community members centered on questions about comfort and interest in using USSD, experiences with intermediation, and discussions on future intermediation modalities.Each FG began by soliciting examples of ICTs and DFS technologies to ensure a common understanding of terms, and we mentioned the more common agricultural extension USSD services available in these villages.
In our FGs with wakalas, we presented the following hypothetical about new ICTs being introduced in a community: Imagine that a government/institution asks you to teach community members how to use a specific new system of mobile technology and encourage them to use it, just as you are now doing for your employers and DFS.Please share how you would want this intermediation to look, and whether you would be interested in participating in this work.How can we design this new modality together?FG questions also addressed running their business, training and organization infrastructure that supports them, and discussions on future intermediation modalities.
Despite the 5-month gap between Phases I and II, the underlying mobile money ecosystem remained comparable during this time, with no new mobile money players involved.Table 1 presents details about our FG samples.This work is covered by the University of Washington's IRB.

Analysis
All surveys and FGs were transcribed and translated into English by two Tanzanian enumerators with a combined 30 years of research and translation experience.FG transcripts were analyzed by the first author, who developed a codebook both inductively and deductively.We used Atlas.ti to iteratively code transcripts and analyzed quantitative data using Python.To preserve the privacy of FG participants, we use W# (e.g., W10) to refer to specific wakala participants and C# (e.g., C12) to refer to specific community participants in the remainder of this article.

FINDINGS
Our research aims to uncover the current state of wakalan intermediation services and the potential to translate their knowledge and augment their labor and wages through intermediation for new ICTs.Below, we present results from our Phase 1 pilot survey with community members and our Phase II FGs with both community members and wakalas.

Pilot Survey Results
Below, we summarize data from our preliminary quantitative survey on the use of USSD services (i.e., a protocol like SMS used to send text messages by many DFSs and ICTs).The survey was intended to identify how community members leverage wakalas to support their use of technology.Our sample of 47 respondents included 19 women and 28 men, with a median age of 38.6, a youngest age of 18, and an oldest age of 74.Twenty-seven percent currently have access to a working smartphone, and the remaining 73% have a feature phone.
Eighty-three percent of survey respondents (39/47) visit a wakala at least once a month to deposit or withdraw money, with a median of two visits per month and average of 1.8 (1 min, 3 max).Thirty-six percent of respondents (17/47) visit wakalas for help with their phones, with a median of one visit per month and an average of 1.4 visits per month (1 min, 3 max).Known uses for USSD apps were money transfers (31/47), buying airtime (15/47), withdrawing funds (11/47), making payments (6/47), and betting (1/47).Figure 1 shows respondent comfort levels with USSD.
Our survey indicates that a significant population (36% of our sample) continues to rely on wakalas for intermediation, and many visit wakalas monthly to conduct financial transactions (83% of our sample).This underscores the importance of intermediation and local intermediators in our studied villages.Further, most respondents needed help using technology at least half of the time or experienced difficulties with new technologies (Figure 1).Therefore, we decided to conduct FGs to probe the human infrastructure intricacies underlying such technology use.

Current Practices and Norms of DFS Intermediation
We describe four types of relationships extracted from themes we generated during our FG iterative coding: community and technology relationship (community interest in ICTs and who needs intermediation to access them), community and non-wakala intermediary relationship (alternate intermediaries the community uses), community and wakala relationship (how intermediation occurs and the change in trust dynamics given organizational policies), and wakala and organizational relationship (challenges to human infrastructure stability).Finally, we address how the community and wakalas envision intermediation if it were structured to serve all involved parties more equitably.All wakala FGs mixed freelance and mobile money wakalas (explained further below).

Community and Technology
Relationship.Generally, participants related their belief that technology had the power to make their lives easier once they understood how to use it, and the vast majority demonstrated an interest in ICTs.For example, C22 mentions, "If it's late and you want to buy electricity credit and shops are closed but you have money in your phone, you can just buy it right there at home with LUKU, and the electricity will stay on." C16 notes that "M-Pesa saves the time of walking toward a bank or anything; you can just do it right where you are.I really enjoy it." Even those who did not actively use digital technology acknowledged that it has changed the landscape of Kagera.
When asked what technologies they were using, FG respondents mentioned few services other than the most common DFS (M-Pesa, Tigo Pesa, Airtel Money) and mobile payment services such as LUKU.C30 notes that awareness of ICTs distributed by the government was limited: "Most people know about Facebook and WhatsApp, but when you talk about other services, they don't know them." We highlight some notable exceptions.One FG member, C13, knew of USSD codes for microfinance banking companies that provided solar equipment loans but could not share precise ones.C19 mentions that Islamic institutions have given him codes to donate to infrastructure projects in different centers: "They might ask for help to build a mosque, so they also send those codes for us to make payments." Learning to use new technologies and ICTs came to community members with time but without formal training.Many participants mentioned teaching themselves how to interact with new technology.For example, C6 states: "Those days, new phones were given with a manual, so I learned most of the operations through that." This approach was shared regarding learning and use of new technology: C16 stated that "Yes, I can [use new ICTs]; with a clear understanding of the previous systems, new systems can't be a problem, " and C15 noted that "Recently, I have gained experience with this service compared to the past, so with a new system, I will continue to learn it a little bit step by step." However, others mentioned that the lack of ICT "training wheels" provided hindered their ability to adopt them; per C9: "Buying airtime is challenging; there isn't enough information concerning the services." We probed which people often seek intermediation, eliciting the perspectives of wakalas and community members.Wakalas characterized people requiring intermediation by first mentioning gender; male and female wakalas both stated that more women came to them for help.Some perceived this as an innate difference in capacity, as W14 notes: "Women are carrying a lot of problems in their head nowadays, so it's easy for them to mix or forget even the simple process of withdrawing the cash from their phones.Men's heads are quick at adopting and managing things." Our qualitative results, in contrast (see Figure 1) show that women were actually less likely to self assess as completely uncomfortable with USSD.Some participants externalized the gender gap as a result of the technology itself; for example, W5 states: "One hundred percent of the things that are in this phone favor men and not women....It doesn't make sense for a grown woman like me to be playing with my phone, but most men do that." Further, per W6: "Women spend their day doing household chores, so they don't have time for that." However, many FG members mentioned that educated women were likely to be able to use technology without assistance.
Age was another factor common in those who sought help.Per W24: "Youth who do not have enough knowledge on how to operate their phones.We still have youth who cannot read and write, too." This was underscored in the qualitative findings (see Figure 1), which show variance in the correlation between age and comfort with USSD technology.However, adults often rely on their children to ask questions and get things done on their behalf.As W38 notes: "They send kids to us with a memo of instruction." Older people with lower vision were cited as often being unable to access technologies on their phone without help.
Similarly, populations with lower literacy were characterized as needing intermediation.Relatedly, an understanding of Swahili or English alone was not always sufficient.Per W4, "What I know is a person may be educated and know English, but the language used in the phones can be a challenge for him/her to understand." Rural areas with differing dialects posed another challenge, as W20 reports: "Yes, because we have a lot of foreigners in this area, so most of them can't use our services, so they come to us for help." Overall, there is consensus among wakala and community members that a lack of agency or understanding is not limited to one type of person; diverse people request help with different aspects of technology.Per W27: "Yes, regardless of age, gender, occupation and financial status when it comes to helping others because people of all kinds may have problems using phones." Community FGs underscored this heterogeneity, with self assessments of the ability to use M-Pesa and similar common USSD applications by oneself ranging from 20-70%.
A minority of community members felt complete agency toward using technology, and the remaining state that they require help from time to time when using their phones.Some respondents opted out of tech even with intermediation due to their lack of understanding; e.g., C23 shares: "I asked for help when I didn't have my own phone, but now I have it, but still I don't ask for help because I don't use my phone often firstly...because I don't have much use for it and secondly I don't have a lot of knowledge about it." However, interaction with new technology was sufficiently daunting for some who could potentially benefit from it to avoid using it; per C14: "I have never asked for such help [...], and I don't use a smartphone either because it has a lot of things hard for me to take unlike a normal button phone" and W10: "There are others [community members] who are just afraid to withdraw money on their own." Some peoples' lack of agency in using the technology is due to infrastructure limitations rather than difficulty in understanding the technology itself.C6 states that: "There are challenges, especially during a bad network; you may buy credit several times without response, and when things are back again, you find that you have purchased triple times." C23 notes: "While I am in the middle of the services, the codes are stuck, then it becomes very hard for me to get the service.They are completely stuck, even if you want to withdraw or to send money from your phone, you can't." This shows that exogenous infrastructural factors beyond challenges with individuals' relationship with ICTs continue to fuel the need for intermediation.

Community and Non-Wakala Intermediary Relationship.
Because there is often no training and coaching, community members seek help from a variety of human relationships.We next review their use of friends and family, customer service hotlines, and government officials.
Friends and family members are the first point of contact for many who experience trouble with ICTs.All FG participants mentioned at least one such trusted individual assisting them.For example, C23 states that: "My child wanted to go to the university, so I asked my neighbor for help, and he use [sic] his phone to search for the joining instruction form in that university website." Silivia reports, "When I bought my first phone, my brother taught me how to use a phone in general, such as how to write and send messages and to make phone calls, serving contacts, looking for contact and so on." However, the intermediation that such others could provide was constrained by the ICT's complexity.Many reached out to customer service hotlines for help, which was effective at times when family fell short.As C6 notes: "For example, [with] Vodacom you dial 100, then you get help, like reversing a wrong transaction or in case you have blocked a sim card." However, talking to someone over the phone had its own limitations.Per C15: "Understanding customer service is around 50% because talking to some[one] on phone differs from face to face talk; you cannot see who you are talking to, and always they have faster conversation than what I prepare myself to respond to." The existence of customer service and telephone-based feedback also opens doors for scams under the guise of technology intermediation.Per C13: "...sometimes you may receive a call from someone pretending to be M-Pesa customer care (or any network agent) giving...false news about winning a prize for your excellence on using M-Pesa or topping up credit, but for you to receive your prize he asks you to carry out several procedures and change your M-Pesa language from Swahili to English.You end up sending him money.So the thing is if your phone operations are set in Swahili and someone [asks you about] changing to English means he/she wants you to press the buttons without knowing what you are doing." Although the Village Extension Officer (VEO) and their team are trusted and educated members of the community, most community members felt like they could not ask them for technologyrelated help.In one FG, everyone burst into laughter when asked if this was a feasible option.When asked about their laughter, C26 responded that it was "because the work of VEO is very different, so in terms of phones, he absolutely can't help us." When asked who had tried to get help from local government officials about his/her phone, all respondents said "No one." Only in cases of ICT-based fraud was it deemed acceptable to contact government officials for help with technology.Per C6: "I can go [to the VEO] if I pick up a phone somewhere to surrender it to him, or if I have received any threat through the phone." C2 notes that "Personally, I cannot ask a ward executive for any help on the phone, whether mechanical or operationally, because a ward executive is not there for such tasks; maybe in case of lost or theft of phone you can seek their help.But issues like how to access a certain service on the phone and so on, I can't personally involve them." Notably, trust in the VEO extended to trust in the agents that they recommended; C25 notes: "I think...using VEO will also help us to know that the agents can be trusted.It will prevent the issue of illegal SIM card registration." We were told of only one successful instance of someone soliciting ICT-related help from government officials.C19 volunteers that: "Yes, there was information I was looking for about a certain university, so I went to the government office, and the officer helped me to write certain codes, and the information I was looking for appears."

Wakala and Community Relationship.
In every village we visited, wakalas were a necessary cog in the finance networks.C22 shares that "you can't put money in your phone without the help of an agent; maybe in case the money is already in my phone and I want to send it to someone else that I can do without going to the agent, but if I have them in cash without the agent, I can't send them." Hence, all people in the village visit wakalas, even if not for intermediation services.
FG participants described two main types of wakalas, each serving a different role: (1) mobile money agents, and (2) freelance SIM card registration personnel.Per C11: "The difference is the [mobile money] agents have an office, but the ones who register SIM cards are freelancer agents who move from one place to another." Mobile money agents are paid a commission when they conduct mobile finance transactions on behalf of customers, and freelancers are paid upon reaching targets of selling mobile cards, vouchers, and so on.
For many respondents, agents are the main, if not only, intermediary for technology.C30 notes that "I have never asked for help from anyone else except the agent, " while C11 states that "I went to the agent because he is the one who knows about phones." When asked why friends and family members could not help, C11 remarked that "even they ask help from the agent." For those needing intermediation who do not visit wakalas, the issue is often accessibility.Per C25: "[I do not go to the agents] because the place that I live in I can't access the agent easily." For most FG respondents, wakalas introduce new ICTs or features of existing services (to reduce learning challenges).C24 notes: "I asked one agent to help [me] take a loan from M-pawa services.But I asked him to lend me money, but he helped me get a loan from my own phone." Wakalas help community members interact with new technology in many ways, including: unlocking phones, troubleshooting network services, sending and receiving money, onboarding to new USSD services, enrolling and troubleshooting on social media, placing bets, fixing internet connections, using calculators, searching for prices, searching google for solutions to problems and fixing them using the phone.Community members report that wakalas also commonly help them block unwanted callers, check a USSD code for a national identification number, retrieve their passwords, unlock their phones, make online payments, open Gmail accounts, download WhatsApp, and access mobile bank services, such as checking their mPesa balance.
We found confusion about free vs paid services that wakalas provide.Some FG participants report avoiding wakalas due to fear of charges.However, many agree that wakalas offer several free or inexpensive intermediation services.Per C24: "I was registered for a different service on my phone that I used to use, but I forgot the codes of those services....The agent helped resolve that without even asking me for money." C18 notes: "I first went to an agent to ask about pay by M-Pesa service because I heard that there was a means of cashing out money with small deductions." However, from the wakalan perspective, offering free help can interfere with getting paid for workrelated tasks.W31 shares: "There are services that customers have to pay for, like swapping phone numbers, but when you told a customer payments are required, he or she doesn't understand." Wakala FG participants noted that most of the "free" tasks they perform could be done by community members themselves.Per W4: "When a person accidentally changes his/her profile without him/her knowing what to do, he/she will come and ask me while it was something he could do by him/herself.Or a customer may have a phone that can use two SIM cards, but he wouldn't know how to block one SIM card to use the other, so he would came to me for help." Wakalas negotiate rates and charge people at different rates for those tasks that may not have an established cost, since this keeps people coming to them.Per W28: "Most of people whom we help without care about their financial status are older and disabled people." W21 notes that "Phone setting most of the time I do it for free, but when the work takes longer than expected, I charge for it." W34 adds that "the only free service I provide to the customer is helping with the air time, but other services, like Personal Unlocking Key issues and resetting passwords, we negotiate." In general, even those with a solid understanding of technology visit wakalas for help with higher order problems.Per C19: "[If] I think that's something that can't be avoidable in terms of mobile network, I might go twice a month, and at those time I might have ruined something in the system, so I will have no choice but to go to the agent." When asked if he went to the agent only to fix things and not to send/withdraw money, for example, C19 replied: "Those are the things I can do myself through the phone."

How Help Happens.
How intermediation occurs between the wakala and community members involves several decisions on the wakalas' part.Intermediation occurs either on the user's or wakala's phone, through teaching the customer, or by acting on the customer's behalf.For example, per C23: "[The wakala] didn't teach me, but he helped me." Acting on behalf of the customer meant that several community members had no idea what the wakalas were doing with their phones, as C26 notes: "He just asked me if I have the card.I showed it to him.I don't know where he clicked, and suddenly my phone was working." C6 adds that "when he returned my phone, I found my SIM card unlocked.So I can't explain what exactly happened." Wakalas report that they decide whom to teach versus on whose behalf to work.Per wakala W13: "It always depends on looking at the age and appearance of a person.I can decide to assist the person directly by taking his/her phone, or I can direct him or her on what should be done." W33 shares that "if they don't know how to read or write, teaching them won't help." Some wakalas prefer to err on the side of teaching, as W17 reports: "If that person is somehow knowledgeable, I always assist them directly from their phones, or otherwise I am always fixing problems by myself." W32 adds that "the services that doesn't [sic] require a customer to use agents work I normally teach them so that next time when it happened [sic], the customer will be able to solve without my help, and also it gives them a chance to help other people who will be facing that same problem in the future." Customers are also sensitive to being taught or having tasks performed for them.As C21 observes: "...I was the one holding [my phone], he only directed me so that in upcoming days, I would be able to do it by myself.And I wrote it down so that when I need it one day, I don't have to bother him anymore." W24, an agent, notes that "I normally position myself closer to the customer so he/she will also see what I am doing, and in case...any question needed to [be] asked, he/she will be able to answer it, " while some leave it up to the customer.Per W14: "It's the decision of the customer; some are interested in knowing, and others do not care; as long as the problem is solved, it's okay with them."

Trust.
We next examine trust dynamics between customers and wakalas through their positive and negative experiences seeking intermediation.We start with those community members who mention fully trusting mobile money wakalas.Per C16: "I have never heard of someone complaining about them, and personally I have always got what I wanted from them.... Mostly they give us a simple and cost effective way of cashing out or sending money, like pay by M-Pesa." This trust might be enhanced by the wakalas' role in providing small or micro loans to community members, whereby they share some risk.Per C23: "I only trust those MPESA agents where I can ask them for some cash, let [sic] say 20,000/=, they lend it to me, then the next day I return it to them." Such shared risk extends beyond established agent/customer relationships.C19 notes: "I went to a certain MPESA agent shop, and I asked him to lend me some cash so that I can pay certain people who were waiting for me, but he asked how can I lend you money while we don't know each other.How will I get it back?But I told him you just give me your number and the money will be sent to you; so he gave me the cash.He gave me 25,000/= I made payment to those people, and I sent his number to a person who sent him money." Bidirectional trust is an established mechanism for daily transactions.Per C16: "I also had help from one agent for my kid's school fees; I went to pay for school fees with insufficient amount, but an agent lent me some, and I repaid the next day." Occasionally, trust arises from helping community members purchase an item from their shop.As C6 relates: "When we don't have enough amount to cover the cost of buying a phone from them, they can accept partial payment and allow you to bring the remaining on the other day you agreed with, but only if you are closely related." From the wakalan perspective, it is challenging to do business without providing loans or supplementary, free intermediation.Per Jaston: "I offer loan[s] to the customer I trust because sometimes it is hard to do business without offering a loan to about 60% of customers; but they pay it back at the end of the month after they receive their salaries, and some other customer may not pay it back.That is very challenging." This demonstrates the material risks that wakalas assume.
However, negative experiences in these relationships demonstrates the fragility of trust.For example, customers use their national ID number to register for SIM cards and get initial phone-related support, and freelance wakalas walk around the village to service those who are not serviced by stationary wakalas in the town.However, mistrust arises, because freelancers have been known to reuse their customers' national id numbers several times and therefore sign a single customer up with multiple phone numbers rather than the single one that they requested.Per C24: "What I am suggesting is, like 70% of these agents can't be trusted.They are hungry for money, and it's the hardship of life." Some wakalas mention that this practice might be due to company-designed pay and incentive structures (discussed below), which often conflict with goals of establishing consumer trust.Per C26: "Sometimes, you may be called by an unknown number and told to send money to that number, but you don't know that person at all." C25 adds: "...multiple SIM cards can be registered [to your national id number] without you knowing, " and C21 states that: "You may have lost your phone and all your registered SIM cards, but when you go to them and tell them I want a registered SIM card, they will give you one without even using your national ID number to register it." In these cases, freelance wakalas often cannot be tracked down due to their constant mobility.Per C12: "Because these [freelancers] have no permanent office, you can go to raise your claim against them, but they are confident to do anything unlike other agents with a physical shop." This feeling is echoed by freelance wakalas themselves, who face the negative consequences of loss of customer trust; per W29, "I am always moving from one market to another trying to find customers.There is a group of people using this opportunity improperly, so it's hard now getting a customer." Such lack of trust has led to violence against wakalas in some areas, as W9 states: "There are others who are not trustworthy; for instance, we went to a certain place wearing M-Pawa T-shirts; the people of that place almost beat us up, because there were some other freelance wakalas who stole from them.But we explained ourselves until they understood us." The relationship of intermediation is one where trust is necessary due to the sensitive nature of the transactions and information that wakalas glean from customers.Per C24: "If someone wants to withdraw money from an agent, they can do all the process by himself near the wakala.But if the person can't write or read, then the wakala will ask that person for some information, like what is your password and the amount you want to withdraw, etc. " Further, the lack of understanding surrounding passwords causes some customers to inappropriately disclose their information; per W18: "Seventy-five percent of customers can't create their own password.You help him create the password, and later on you teach him how to create one himself so that he can change it.So is it possible they allow you to create a password for them not because they trust you, but just because they don't know how to do." This was echoed several times by community members.It is clear how passwords could be misused by wakalas, but their customers' naivete can lead to negative consequences for the wakalas, as well.Per W33: "What I do to make sure that I continue to maintain the trust I have built is to not help customers with password issues unless that customer is my loyal customer whom I know for sure doesn't get service from other agents, and he or she is old or can't see; but for every other customer, I don't do that because that has caused countless problems, and it may lead...the trust you have built to die." The inability to see who is misusing a password may thus lead to a loss of trust and therefore of customers.
It is evident that customer trust in wakalas is therefore directly linked to wakala livelihood.Wakalas use proxies to measure customers' trust and loyalty to ascertain how well they are doing in a given month.Per W25: "I know that a customer trusts me when they call me and ask me about other services [I offer] and if there is any change he should be aware of." W39 adds that he knows he is trusted, "because some of them [customers] leave their money with me so that I could send it to them when I have credit in my phone." Many wakalas report that customers form loyalties with one wakala in the area; W35 states that "they are loyal to me; when I am not around, they will wait until I return to get services." How the wakalas foster trust varies.However, each employs intermediation strategies to keep customers coming back for paid services.Some wakalas follow up with customers; W27 notes: "I normally take my customers' number so after a month I call them, and they are still using the same SIM card I sold to him; so that gave me the sense that my customer trusts me."Others rely on their uniforms, proximity to customers and working daily in the same places as a means to maintain customer loyalty.Some offer financial discounts, like W35: "[I] sometimes provide the services with a discount price" to gain customers.Some cite a common language as a means to reach customers who may not otherwise enter the system; W9 notes: "The language I use to communicate with them helps them trust me."

5.2.4
The Wakala Infrastructure.Wakalas work within a complex infrastructure that includes their mobile network operators (MNOs).We next describe their experiences becoming and serving as intermediaries.
Training experiences to prepare themselves to become intermediaries varied greatly.Formal training was through seminars, as W17 notes: "These companies have been providing different seminars on how to provide this service.But those seminars are very few.I don't even remember the last time I attended." They also learn through shadowing an existing employee.Per W33: "Before, it wasn't easy to get this job.You needed to be at least a form four graduate every time we went to the field so that he could teach me how to work." Agents also reported learning their trade through phone conversations, as W20 notes: "Through phone calls and sometimes face to face when I failed to understand some things." Informal training was more common, since some wakalas acted on behalf of their employers, who owned the kiosk and were formally registered.Per W35: "I taught myself; I observed how others worked." W5 shares: "I was taught by my daughter; she taught me how to send and withdraw money." W4 adds: "My boss taught me." As with any technology, all ICTs that the wakalas intermediate for change from time to time, sometimes without clear notice given to wakalas or community members.Many wakalas expressed challenges in informing customers of these ICT changes due to the lack of communication from the MNOs to the customers.This impacts customer trust in their services.For example, per W13: "When one customer tells others that I am reducing their money, then they stop coming to my shop.This is due to lack of education on the withdrawal costs introduced by the government.These telecommunication companies like Vodacom or Tigo can solve this problem by providing education." W11 adds: "Media advertisement...on increase of charges done by the government should start as soon as the day the announcement is given out." In addition to straining trust relationships, the increase in charges also directly impacts a wakala's ability to solicit new customers.Per W9: "The more charges there are, the less likely a customer is to come to the agent for support." In the same vein, as wakalas are the face of the technology to the community, if the technology fails, some blame the wakala.W24 notes that: "I may send money to the customer, and he doesn't check if the money is in his account; then, after two to three days, he returns to me and said I didn't see the money.I will then have to call the headquarter to ask for that money, but the charges of sending that money to the customer account will be deducted from my pay." Further, there is a lack of education regarding the ecosystem of scam messages and how to differentiate real from fake messages sent by the MNO.For example, per W32: "Sometimes, a customer can receive a scam message that shows him or her that he or she received money and asks you to help him or her to withdraw the money, but when you try to do that you find out he or she doesn't have money in the account, so when you told him...it is hard for him to understand what is a scam....It once caused a customer to report an agent to the police station." To our participants' knowledge, there is little support for ICT-related fraud via police or other authorities.W11 notes that "there are many thieves with new techniques, and when reported to the police, it can even take you seven months making follow-up on the same matter; we are losing our money." W23 adds: "When I register a customer SIM card, and another agent unregister[s] it, it's very challenging, " as this registration would fail to count toward his quota.
Given economic vulnerabilities due to scams and unethical practices, wakala compensation is low, especially for freelancers.Per W33: "You may walk a long distance to find a customer, but still payments are very low." For those using transport to reach more distant customers, agents mention that this consumes a significant portion of their earnings.Per W27: "The day of payment, he or she will just be paid the amount he took from his own pocket." Compensation for freelancers occurs after they achieve their MSN-derived targets; therefore, financial insecurity is high until they reach that goal.For mobile money agents working for more than one MNO in our groups, the salary was about TSh 200,000-400,000/month (USD 100-200) for working 50 h a week.Further, wakalas are not compensated for many forms of intermediation they provide to retain customers; though this is really a form of advocacy for their MNOs, they do not receive the same level of support in return.Some wakalas use systems to give feedback to their employers about challenges they or their community members face in using their technology.Some simply call customer service lines for their company, like W19: "Yes, we have numbers for reaching out [to] customer cares in case of any problem." Some call or visit their superiors, and a select few have morning meetings before starting work to talk with employers about the challenges they faced the day before.One agent mentioned that they use a WhatsApp group to communicate feedback to employers.While some believe their feedback makes a difference to the technology, like W18 ("Yes, because I believe challenges reported to my bosses are being shared to the top management and lastly to the government."), the vast majority know of no mechanism to give feedback to their employers.As W21 remarks, "Personally, there is no system for me to give feedback."

Future Intermediation Possibilities
We asked wakalas and community members to share their perspectives on how they envision changes to intermediation if the government were interested in providing community-based support for new ICTs.

Community Thoughts on Extending Intermediation.
All community members agreed on the need to increase intermediation support to improve access to digital tools, even if they manage to use technology by themselves.Per C24: "Few of us manage to do it alone, but the Government started the adult school; they can also start something like that which will be able to educate people on phones." Per C5: "People to train us are important because we differ in terms of understanding and of understanding each other, but having specific agents where you can direct people for better understanding sounds good." Per C15: "I will be happy if they come to teach us because there are things I don't know.To me, I think the availability of the people to train us will be able to learn a lot of things at a time.Let people come to train us and also keep instruction in the system itself." Some felt that having more mobile agents (freelancers) would help reach remote areas where understanding and use of services is low.Per C8: "I think having an agent will be helpful, especially the freelancer ones because they will be able to reach even the people who won't be able to go to town and get help.To understand something, people must be there to guide us.When you send a child to school, a teacher should be there to teach a kid to understand, so there should be people to teach us a new system, too."

Wakalas' Thoughts on Employment
Stability.We asked wakalas how likely they were to stay in their current role.Some mentioned not staying for more than a year; per W28: "I personally don't see myself staying longer in this job.There are a lot of challenges.We don't have contracts, so we are being paid less but you have worked hard." W32 adds: "Right now if I get a better job, I won't do this anymore." Some described how they supplement their wakala job income with other work due to a lack of customers (see Figure 2(a)); per W5: "We aren't that busy.It gives us time to do other things; apart from the agents' work, there is agriculture." The level of job stability agents felt is influenced by several factors.W4 notes that "it depends on the area you are working at, " while W3 observes that "it depends on the customers you have at the moment." Only a few agreed that they are likely to stay in the role for another 5 years.However, most shared that they would be interested in participating in additional work similar to that they already do, with the following caveats: (1) better compensation (W3: "...I shouldn't do [more intermediation] if there is no profit in it."), (2) localization (W21: "If I can work from my office, it won't be a problem."), (3) time (W39: "It depends on if it will contradict my working hours or not.I don't want any confusion to arise between my current work and that."), and (4) new customers (W7: "This has became a district, and population is increasing so [additional responsibilities re intermediation] will be an opportunity to get customers.") However, some wakalas mentioned that they are too busy to take on additional work beyond their current jobs; per W37: "Based on what I am doing, I can't add something else to it because it needs attention.Even if I will be at home, my work continues because customers come to me."

New Modalities of Intermediation. Most wakalas prefer a hybrid training model if they
were to be taught how to use new ICTs, i.e., a combination of in-person training and online support.Solely online training was not the preference of several wakalas, who noted that it is difficult to understand and follow instructions without a person who can clarify questions.Further, they recommended creating WhatsApp groups for follow-up after training ends in case they encounter problems.They also recommended compensating for transport to/from training.
At a higher level, they recommended involving the community from the start.Per W26: "I think it will be good to make an announcement in the community that certain agents will be providing certain services before we start talking to the community people." To this end, allying with the government was suggested during training and beyond, as W27 shares: "While providing the training, we should be with government officers so that when the training is over and we start to talk to people in the community...government officers will be able to identify us." This was part of how wakalas wanted to foster and maintain community trust.
If they were to disseminate new ICTs, then agents would like to see a number of changes made.First, they thought there should be a robust channel of communication to customers directly from ICT creators to prevent them from being the only face of the technology and risking loss of trust when tech updates were made.Per W1: "People need education on the charges they should expect and how the technology should function.[ICT creators] can provide seminars, advertise in different media." In tandem with this was the design of safeguards in the technology to support intermediation activities; per W10, "Introduce a special application which will be used on correcting all challenges and other information." From both wakala and community perspectives, the need to couple intermediation with the availability of easy instructions in the app or online was raised.C9 noted: "I prefer the agents to be available, but I would also advise that instruction of that service should be available in Google because now the number of people using smartphones is larger." Wakalas also suggested a customer service line that could operate when they were not available.
In addition to intermediating for existing ICT users, wakalas shared that they could promote new ICTs to customers that seemed likely to benefit from them.This was already part of their current work; as W14 mentions, "Yes, I do always tell customers about the availability of useful programs found in their phones." Agents seemed to conduct such promotions especially when new features favor customers financially, even if that means that customers are less likely to come to them for subsequent transactions with physical cash.For example, per W25: "I encourage them to use mobile payment systems like buying LUKU and paying different bills through [the] phone if they are not doing so." W4 adds: "I encourage the customer to use [a] new system when that system happens to be cheaper than the previous one." Sometimes, new product promotion occurred at the will of the company.W33 notes that "we only emphasize [promotions to] customers based on what we have been told by the upper authority.Right now, our main emphasis is to tell customers about the shift from 4G to 5G Vodacom network." We asked wakalas about compensation for intermediating new ICTs.Two prominent models emerged based on their work experiences.The first was a commission-based method through which they would be paid for meeting targets or based on their level of engagement.Per W40: "I would like to be paid according to my work performance." For this method, agents would want to be told how much they had accrued in terms of salary at the end of every work day.The second was a fixed monthly payment; per W35: "I prefer contract payment because it gives you an assurance of being paid compared to commission." In addition to monetary preferences, agents also requested phones, company identification, and uniforms.Overall, wakalas felt that their jobs would be easier if the surrounding tech infrastructure were strengthened.W35 observes that "[the] main challenge is [the] network.Some customers may need a certain service but at a time the network won't be working." They requested that the technology infrastructure be improved to incentivize them to conduct further intermediation.

DISCUSSION
A summary of our results is in Table 2, we now draw on focus group findings (Section 5) and intermediation models (Section 3.4) to present three key factors that influence the potential of leveraging the Tanzanian wakalas to intermediate other ICTs: (1) incentive alignment and the limits of pro bono actions, (2) appropriate training and support infrastructure, and (3) the need to maintain trust and build reciprocity.

Incentive Alignment and the Limits of Pro Bono Actions
In our FGs, wakalas asked to be compensated appropriately for their intermediation work.The wakalas discussed how the work they do extends beyond direct intermediation of a technology platform to include other, uncompensated activities.Family/friends, customer service hotlines, and Govt are relied on for ICT support in differing degrees and with limited impact.

Wakala and Community Relationship
From loan services to password management, wakalas provide a host of uncompensated support beyond digital banking services.Scams and ICT breakdowns negatively impact customers' trust in wakalas.Wakalas use their often uncompensated loan service as a mechanism to build trust with customers.

Wakalas and Their Organizations Relationship
Limited training for wakalas complicates the work of intermediating software changes for customers.A target-based compensation model creates a perverse incentive structure for wakalas.

Community Thoughts on Extending Intermediation
Community members are interested in expanding intermediation to more ICTs.Freelancers are seen as potentially having more impact in hard to reach areas.

Wakalas' Thoughts on Employment Stability
Some wakalas want additional stability and compensation.Others want more opportunity for well-compensated intermediation work, more structure in their work, and the ability to interact with new customers.
Research in economics shows that the design of incentive-compatible compensation schemes is a difficult task, especially in the presence of strategic agents [23,43,72].Reference [39] discusses the evolution of incentives schemes used by Mobile Vaani3 to compensate and reward the work of its volunteer network.As discussed in Section 3.4, changes in Mobile Vaani's incentive scheme were influenced by changes in the intermediation model and decisions about the level of formality and accountability needed for Mobile Vaani intermediaries.
As Mobile Vaani compensation shows, designing appropriate incentive mechanisms for ICTD technologies is a balancing act that requires careful attention to social and monetary incentives.For example, the authors of Reference [65] conducted a study to identify the reaction of community health workers (CHW) in five low-and middle-income countries to different incentive structures.According to the study: "The issue of payment is more complex than might initially appear and can undermine the programme if not thought through well.If payment or incentives are perceived as inadequate or do not flow in a timely fashion, then this can be demotivating to the CHW.Part-time models should not place unrealistic expectations on the CHW's time or capacity.Payment or incentives that are perceived as inadequate or do not flow in a timely fashion can be demotivating to the CHW.Part-time models should not place unrealistic expectations on the CHW's time or capacity." The authors discuss how an undue emphasis on incentives can "create[s] an environment where intrinsic motivation can be lost" and "focus placed on incentives" [65].They call such environments hybrid markets, where a mix of monetary and social markets is required to sustain an intervention.
From our FGs, we also see these "hybrid markets" as wakalas describe aspects of their work: the financially compensated tech support of mobile payment apps and uncompensated work they do to gain and maintain customer trust.When asked about the type of incentive/compensation model they would like to see, some wakalas wanted a performance-based model that tied payments to their ability to deliver good and effective services, while others wanted an assurance-based model that provided a guaranteed income irrespective of performance.
The effective design of any incentive scheme for the wakalas will need to balance the requirements of income stability and fair effort-based pay.However, given the challenges of intermediation described in Section 5, care needs to be taken when designing partially financial and partially social incentives mechanisms, since they may interact in dynamic and counter-intuitive ways, as evidenced in the case of Mobile Vaani.Like Mobile Vaani, many ICTs will also require continuous changes in their incentive mechanisms and chosen model of intermediation: volunteers versus a highly structured and intentionally funded community.The ultimate success of this activity will be a function of whether there is a market for their services.

Appropriate Training and Support Infrastructure
In our FGs, wakalas expressed the need for better in-person introductory and on-going training from the companies/managers of the technologies they intermediate.They also stressed the need for more shared risk approaches to intermediation that could prevent reputational damage when customers have issues with technology operation and performance that are beyond their control.Further, they expressed interest in intermediation work that would provide opportunities for in-person training, online support, and more channels for education, and they wanted more structure in how they were organized in relation to each other, customers, and the ICT creators.
In this regard, the models of intermediation used by Mobile Vaani to train its Club Volunteers and by DHIS2 through its in-person DHIS2 Academies may resonate with wakalas.ICTs with a Mobile Vaani-like intermediation model may be more appropriate for wakalas to intermediate, since they can also support more robust engagement with customers beyond merely fixing technical issues and move toward increased literacy and additional ICT uptake.Further research could explore how intermediation structures can facilitate multi-modal approaches to training with strategies specifically for cooperative teaching as well as for acting on behalf of users uninterested or otherwise unable to engage, as studied by Parikh and Ghosh [51].

Tensions in Wakala Trust
Networks.Research defines trust as "a bet about the future contingent action of others" [13] or "the willingness to be vulnerable to the actions of another party that is acting on one's behalf with or without the ability to monitor or control that other party" [24].Combining these definitions, we can state that trust is contingent on the potentially unverifiable actions of an individual or system acting on behalf of the trustee.
In computer science, trust has been been well studied across a number of contexts, from peerto-peer exchange to consumer-to-consumer e-commerce [3,40,54].Generally, there are two notions of trust discussed in the literature: "trust in strangers" and "trust in a platforms/institutions" [13,24,27].
For DFS intermediation, due to the lack of trust in institutions and their ICTs by community members, wakalas provide significant added value to DFS institutions [17].Our findings in Section 5.2.3 discuss the importance of trust and its impacts on the livelihood of wakalas.The overwhelming view from our FGs is that the level of trust customers place in wakalas is contingent on two main factors: (1) whether the wakalas provide loan services and (2) whether they are mobile.The wakalas' provision of loan services is a way to build reciprocity with their customers, thereby increasing the likelihood of continued patronage of their intermediation services.
We found a distinction in customers' trust relationship between freelance wakalas, who provide SIM registration services and are not limited to single geographic areas, and their more fixed counterparts, who primarily offer mobile money services.As discussed in Section 5.2.3.2, customers place less trust in freelancers, chiefly attributable to their mobility and tendency to fraudulently register customers with multiple SIMs.In addition to such fraud, the growth of spam calls and the inability of customers to identify spam may complicate intermediation work: Customers tend to blame wakalas for any discrepancies in their accounts, and password use is impossible to audit.Fraudulent freelancer behaviors and mobile money scams thus threaten the general trust in wakalas, their incomes, and their personal security.This lack of trust is bi-directional; as we discuss in Section 5.2.3, wakalas, particularly those providing loan services, are also seeking ways to assess the trustworthiness of customers.

Mechanisms for Maintaining Trust.
In the absence of mechanisms to protect trust, customers and wakalas are creating proxies to protect and maintain trust.From FGs, we learned that customers and wakalas need mechanisms, both technological and social, to support trust maintenance.We discuss two categories of such mechanisms: those pertaining to the design of products requiring intermediation and those pertaining to external processes to monitor trusted parties.As discussed in Section 5.2.3, customers and wakalas have asked for ways of building technologies that provide mechanisms for trusting each other.
There is tremendous opportunity for ICTD researchers to explore the integration of proven trust and reputation technologies into ICTD technologies in way that supports safe and trustworthy intermediation.For example, ephemeral and auditable credentials that allow agents to access a customer's account without requiring a customer's account credentials [59,63,74,75] can make the work of intermediation more transparent and decrease the likelihood of fraudulent uses of a customer's private information.Another potential trust-maintenance technology is bi-directional reputation systems with strong sybil-proof guarantees [30,42,62] that both customers and wakalas can use to gauge each others' trustworthiness.Considerable research shows that in a digital world, the ability to access "cheap pseudonyms" significantly impacts social trust given the lack of reputational consequences for behaviors [19,20].Reputation systems designed to be sybil-proof and publicly verifiable could serve as deterrents to agent malfeasance [54,55].
However, such mechanisms fundamentally challenge the highly personal nature of trust in the social context where wakalas provide intermediation services.They require a "trust, but verify" mindset that may have perverse consequences on interpersonal trust relationships that exist in these community.As agents intermediate digital services that leverage and scale their interpersonal trust relationships, there is a need to simultaneously investigate what hybrid and culturally appropriate trust and reputation maintenance tools are needed to handle the tensions arising in this setting.

CONCLUSION
We characterized the current practices of wakalas and explored key factors that influence the potential of leveraging the wakala as human infrastructure for other ICTs.Adopting a mixed-methods approach, we used a quantitative survey and FGs with community members and wakalas across five villages in Kagera, Tanzania.Our findings provide evidence that community 3:23 members in Kagera have ongoing challenges accessing technology and require intermediation.Through our focus groups, we discovered current practices of intermediation and explored ways in which these practices could be expanded to the benefit of wakalas, community members, and ICT creators.We also contributed three key factors useful for framing discussions about trade-offs in various models of intermediation: incentive alignment within the limits of altruism, appropriate training and intermediation structure, and the need to maintain trust and build reciprocity.

Fig. 1 .
Fig. 1.Community member responses to the following survey question: "How comfortable are you with using USSD technologies?"
Today, many ICTD stakeholders focus on the development of technologies known as digital public goods (DPGs): these include open-source software, open data, open AI models, open standards, and open content [48].Unlike traditional ICTs, the openness of DPGs may create new market dynamics, since their flexibility and availability engender positive feedback loops and spillover effects.Though beyond our scope here, future work could explore how and in what ways the very open DPGs influence their ability to be sustainably and effectively intermediated by agents like Tanzanian wakalas.

Table 1 .
Focus Group Participation Summary

Table 2 .
Summary of Findings