Citizen sensors or extreme publics? Transparency and accountability interventions on the mobile geoweb

Abstract

Transparency and Accountability (T&A) interventions are emergent social technologies in middle and low-income countries. They bring together citizen sensors, mobile communications, geo-browsers and social organization to raise public awareness on the extent of governance deficits, and monitor government's (in)action. Due to their novelty, almost all we know about the effectiveness of T&A interventions comes from gray literature. Can citizen sensors radically increase the transparency of the state, or are changes brought about by T&A interventions more likely to be incremental? We review the literature on transparency policies and describe their drivers, characteristics and supply–demand dynamics. We discuss promising cases of T&A interventions in East Africa, the empirical focus of an on-going collaborative research program. We conclude that the effect of T&A interventions is more likely to be incremental and mediated by existing organizations and professional users who populate the space between the state and citizens. Two elements at the interface between supply and demand seem rather crucial for designers of T&A interventions: accountability-relevant data and extreme publics.

Keywords:

• citizen sensors
• transparency
• accountability-relevant data
• extreme publics
• mobile geoweb

1. Introduction

Five years after the notions volunteered geographic information (VGI) and citizen sensors first made their debut (Goodchild 2007), VGI research reached a state of maturity that prompted the editors of Crowdsourcing Geographic Information: VGI in theory and practice to contemplate a fourth paradigm for future research (Sui, Goodchild, and Elwood 2012). The editors urge us to study the heterogeneous social relations through which VGI is produced and their implications for participation, politics and collective action. The locus of this study is one of their suggestions for future research: ‘what kinds of state-citizen relationships are produced or transformed through the creation and use of VGI?’ (Elwood, Goodchild, and Sui 2012, 368).

We focus on transparency, a state–citizen relationship that enables citizens to ‘see’ the inside workings of government and to monitor government's actions. To ‘see’ the state, citizens must access the information that government generates as part of its day-to-day activities and that covers most aspects of our lives. Transparency policies make the ‘seeing’ possible. Examples are Freedom of Information (FoI) laws, Targeted Transparency (TT) policies and Open Government Data (OGD) initiatives. Laws for Freedom of Information have a distinguished and long history (Hood and Heald 2006). Targeted transparency policies are regulatory innovations that have percolated through the political system in the past few decades in response to crises and have since become mainstream policy tools (Fung, Graham, and Weil 2007). Open Government Data initiatives are of very recent vintage (Memorandum OMB 2009). They are actively encouraged by the Open Government Partnership, a global initiative aiming at more transparent, effective and accountable governments, with institutions that empower citizens and are responsive to their aspirations (http://www.opengovpartnership.org/). To date 47 governments around the globe, including three in Africa – Tanzania, Kenya and South Africa – have committed to develop OGD initiatives. Tanzania and Kenya delivered their commitment in 2011. One of their commitments is the enactment of a Freedom of Information (FoI) law. Tanzania is currently studying global best practice in FoI legislation in order to generate its own FoI bill, while Kenya is in the process of drafting a FoI law.

In the absence of transparency policies, citizen sensors and volunteered geographic information may come to the rescue (Foresman 2008; Goodchild et al. 2012). The technology for transparency network (http://transparency.globalvoicesonline.org/about), a research and mapping project supported by the Transparency and Accountability Initiative, has documented to-date 63 cases of transparency and accountability (T&A) interventions that crowdsource citizens' reports in middle and low-income countries. In about a third of the cases, citizens' reports are geo-referenced and mediated by the mobile geoweb. Well-known examples of T&A interventions are Uchaguzi, for monitoring Kenya's 2010 constitutional referendum, and Huduma, for monitoring public services in Kenya. Both are follow-ups of Ushahidi, now an influential NGO specializing in developing free and open source software for information collection, visualization and interactive mapping.

In T&A interventions, citizen sensors use diverse technologies for data collection, visualization and social networking. They target different social actors (e.g. the executive, or legislative branch of government), or different functions these actors perform (e.g. elections, budgetary processes, or public service delivery). However, in all cases T&A interventions bring together information technology and social organization to raise public awareness on the extent of governance deficits, and monitor government's (in)action. When they are in place, citizen sensors themselves generate the information to render government's (in)action transparent (Georgiadou, Budhathoki, and Nedovic-Budic 2011), or coproduce it with government (Linders 2012).

Transparency policies and T&A interventions have differences and similarities. They differ in the direction of information flow. When transparency policies are in place, the direction of flow is from the state (or from private entities regulated by the state) to the citizens. In T&A interventions, information flows in the opposite sense – from the citizens to the state. Transparency policies and T&A interventions are similar in that expectations about how they will work in practice, when implemented, are usually overoptimistic. The excessive optimism is due to the rather crude, but widespread assumptions regarding the nature of the public, of the state and of the state-citizen interface. Scholars of transparency have convincingly argued that these assumptions are simplistic, and have advanced more realistic conceptions of the public, the state, and the state–citizen interface (e.g. Fenster 2006, 2012).

However, while the actual workings of transparency policies are well documented in the scholarly literature, almost all we know about the functioning of T&A interventions comes from very recent and gray literature: reports of the Transparency and Accountability Initiative (e.g. Avila et al. 2010; Fung, Russon Gilman, and Shkabatur 2010; McGee and Gaventa 2010); the World Bank (2012), speeches hosted by the Program on Liberation Technology of Stanford University (e.g. Fung 2011) and working papers (e.g. Fung, Russon Gilman, and Shkabatur 2012). Empirically and theoretically grounded research, especially in middle- and low-income countries, is scarce with the exception of Meier's (2011) dissertation, on citizen sensors using the Ushahidi platform for political mobilization in Egypt and Sudan. Thus, there is much to learn from the literature on transparency policies, especially regarding more realistic conceptions of the public, the state, and the state–citizen interface.

In this paper, we ask whether citizen sensors creating and using VGI can radically increase the transparency of the state, or whether changes brought about by T&A interventions are more likely to be incremental. We review and synthesize the scientific literature on relevant transparency policies – Freedom of Information (FoI) laws, Targeted Transparency (TT) policies and Open Government Data (OGD) initiatives. We review the gray literature on T&A interventions and single out promising cases in East Africa, the empirical focus of our research. We draw on interviews with NGOs in Tanzania conducted in May 2012, in the context of a collaborative research program, titled Sensors, Empowerment and Accountability, and funded by the Netherlands Organization for Scientific Research. Ultimately, we expect to learn lessons from the actual functioning of transparency policies and interventions. The lessons will have implications for us and other students and designers of T&A interventions in East Africa, and potentially in other middle and low-income regions.

The rest of the paper is organized as follows. Section 2 reviews those types of transparency policies and interventions that relate the citizens to the state. Sections 3 and 4 highlight characteristics, drivers and demand–supply dynamics of general and sectoral transparency policies and interventions, respectively. Section 5 discusses implications for the design of future T&A interventions and offers conclusions.

2. Types of transparency policies and interventions

Transparency refers to the open access to government information, and allows citizens to ‘see’ and monitor government's actions. Transparency is the informational component of open government, while participation is the interactional component (Meijer, Curtin, and Hillebrandt 2012). Participation refers to open access for citizens to decision making arenas, and allows citizens to ‘speak’ and influence government decisions. When combined, the two components of open government – informational and interactional – are the foundation of informed debates between the governors and the governed. Openness, ‘seeing’ and ‘speaking’, strengthens public accountability, the hallmark of modern democratic governance. Without accountability, ‘democracy remains a paper procedure if those in power cannot be held accountable in public for their acts and omissions, for their decisions, their policies, and their expenditures’ (Bovens 2005, 182). Here, we focus only on transparency, the informational component of open government. Figure 1 shows relevant types of transparency policies and interventions. We include only those types of transparency policies and T&A interventions that make the state visible to the citizens and exclude a variety of other types (Heald 2006) that do not fulfill this criterion.

Figure 1. Types of state-citizen transparency policies and T&A interventions.

We make a first level distinction based on the kind of information (general versus sectoral) released. The first type is ‘access to general information.’ The disclosed information is deemed general in that it does not apply to a specific government sector or policy domain. It includes two sub-types: passive release of information and its recent digital variant, active release of information.

Laws for Freedom of Information (FoI) exemplify passive information release. FoI laws oblige the state to release information to citizens who claim their ‘right to know’ (Fenster 2012). The advent of the internet played a major role in the move from passive to active release of information. The influence is visible in the FOI laws themselves. One of the youngest FOI laws, the Indian Right to Information (RTI) Act of 2005 encourages the use of internet for proactive government information disclosure (Roberts 2010).

Open Government Data (OGD) initiatives epitomize active information release. A major impetus to active release of information was President Obama's Memorandum for the heads of executive departments and agencies in January 2009 (Presidential Memorandum 2009). The Office of Management and Budget (OMB) put into practice the principles of ‘transparency, participation, and collaboration’ advanced by President Obama (Memorandum OMB 2009). OMB required agencies ‘to publish information online […] that can be retrieved, downloaded, indexed, and searched by commonly used web search applications […] in an open format […] that is platform independent, machine-readable, and available to the public.’

The second type is ‘access to sectoral information.’ The disclosed information is sectoral because it applies to a specific policy domain, e.g. environmental risks. It includes two sub-types: targeted and horizontal transparency. Targeted transparency (TT) policies further specific policy objectives, e.g. the reduction of toxic pollution (Fung, Graham, and Weil 2007). They are underpinned by laws, e.g. the Emergency Planning and Community Right-To-Know Act (http://www.epa.gov/oecaagct/lcra.html). TT policies regulate specific private and public agencies, e.g. polluters, and mandate them to disclose publicly, factual information, in standardized, disaggregated, comparable formats, concerning specific products and practices (e.g. amounts of toxic pollution at a location). TT policies started off in the second half of the past century as unplanned inventions by political entrepreneurs responding to crises.

The idea of horizontal transparency, the second sub-type of access to sectoral information, can be traced back to a World Development Report in 2004. The report concluded that the ‘long route’ of traditional accountability (via elected politicians and public officials, service providers through to citizens) was failing the poor (World Bank 2004). The solution was the strengthening of a ‘short route’ to enable citizens to demand social accountability directly from service providers (McGee and Gaventa 2010). Social accountability implies horizontal transparency. Horizontal transparency creates a level playing field between citizens and government. Citizens can ‘see’ the information of service providers, e.g. providers of water and health services, or produce it themselves if the information is not available. Within the sub-type of horizontal transparency, two variants can be further distinguished: Transparency and accountability (T&A) interventions and collaborative T&A interventions.

Transparency and accountability (T&A) interventions are the most commonThey are emergent social technologies in middle and low-income countries. They leverage mobile communications, social media and increasingly the Geoweb. The information disclosed is sectoral (e. g. public health, and education services, water and sanitation services, electoral issues) but often aggregated to guarantee the anonymity of the citizen sensors. There are indications that T&A interventions are becoming genuinely collaborative, as we show in section 4.2, using Kenya's MajiVoice initiative (World Bank 2012).

Before discussing the different transparency types in detail, it may be worth clarifying the different politics that underpin them (Stalder 2011). Freedom of Information (FoI) and Open Government Data (OGD) are underpinned by liberal politics – the demand for transparency is directed at state institutions to create accountability to citizens from whom they derive their legitimacy. Targeted transparency is underpinned by neo-liberal politics – the demand for transparency reduces information uncertainty and information asymmetry in the market. This insight may help us understand better the degree to which transparency policies ‘stick’ in different political contexts, but exploring this aspect further here is beyond the scope of this paper.

3. Access to general information

To better explore the interface between the demand side (citizens) and the supply side of transparency (state) for FOI laws and OGD initiatives, we first summarize their most salient drivers and characteristics (see Table 1).

Table 1. Drivers and characteristics for transparency policies (access to general information).

	Freedom of Information (FoI)	Open Government Data (OGD)
Drivers	Political activism by civil society	Advocacy by information technology champions
Characteristics	Mandatory release	Discretionary release
	Freedom of press	Crowd-sourcing, mashing and hacking
	Accountability via legal rights	Accountability via IT and communications

3.1. Drivers and characteristics

Freedom of information (FoI) laws are characterized by mandatory release of all government records unless specifically exempted by the law itself. Individuals can access them by mobilizing their legal right-to-know, to exact accountability. FoI laws flourish when the freedom of press is guaranteed (Hood 2006). It is no coincidence that the first act for mandatory access to government records passed by Sweden in 1766 assured the freedom of press. Freedom of information and freedom of the press are intimately linked (Nam 2012). ‘Users of information must have access to the widest possible diversity of points of view on a particular issue to hold the government accountable’ (Ackerman and Sandoval-Ballesteros 2006, 89). Political activism and advocacy are the main drivers for enacting FoI laws. They play a crucial role in passing, implementing, and enforcing FOI, but also in keeping FOI regulation intact once it exists. For example, attempts by the Indian government to restrict the Indian Rights to Information (RTI) Act failed due to pressure from civil society groups (Raman 2012; Roberts 2010).

Open Government Data (OGD) initiatives entail the active release by government of raw, machine-readable, high-value information in open formats. High-value information is defined as ‘information that can be used to increase agency accountability and responsiveness; improve public knowledge of the agency and its operations; further the core mission of the agency; create economic opportunity; or respond to need and demand as identified through public consultation’ (Memorandum OMB 2009, 7). OGD initiatives take a laissez-faire approach. Government passes on to individual agencies the responsibility of making open government plans and deciding on the data to be released. Agencies thus have the discretion to decide which previously unpublished ‘high-value’ raw data they can place on a national Data.gov site (Shkabatur 2012b).

Further characteristics are technology-mediated accountability, and a crowd-sourcing vision. Both are instrumental for channeling crowd knowledge in the policy process and materializing the participatory principle of the Open Government Directive, which states that ‘[p]ublic engagement enhances the Government's effectiveness and improves the quality of its decisions. Knowledge is widely dispersed in society, and public officials benefit from having access to that dispersed knowledge’ (Presidential Memorandum 2009, 1). Agencies simply publish raw machine-readable data. The public decides how to use or hack them and for what purposes and how to mash them with other data (Brito 2007). Advocacy by information technology (IT) champions is the main driving force (Fenster 2012).

3.2. The demand and supply of transparency

Critics are impatient with the dashed expectations of FoI laws (Roberts 2006; McDonald 2006; Fenster 2006) and the progress of OGD. They complain that changes so far have been incremental rather than revolutionary (Fenster 2012), that implementation so far has occasionally been ineffective (Lakhani, Austin, and Li 2010), and that little is known about the effect of transparency on the relationship between institutional stakeholders and the public (Scholl and Luna-Reyes 2011). They claim that online data is often not in useful formats (Brito 2007) and does not foster public accountability when available (Shkabatur 2012b). The frustrations of the critics may be due to three widespread but rather simplistic assumptions regarding how FoI laws and OGD initiatives work (Fenster 2006): First, the public – demand side of transparency – is assumed to be a homogenous collective that awaits disclosure and will react to it in predictable, rational, and informed ways. Second, the interface between the public and the government is assumed to be unproblematic. And third, the government – supply side of transparency – is assumed to be an uncomplicated transmitter of information, if the proper disclosure requirements are in place. In reality, the three assumptions do not hold. We will examine each one in turn. Table 2 summarizes pragmatic propositions to overcome the limitations of transparency policies.

Table 2. Pragmatic propositions for transparency policies (access to general information).

	Limitations	Propositions
Demand-side for transparency	Divided public	Mobile communications as technologies of choice
	Only organized and technology-savvy public demands transparency	Mobilization of ‘info-mediaries’ (activists, journalists, civic hackers)
Interface between supply and demand	The state can be unresponsive to the demands for transparency	OGD initiatives in dialog with civil society for specific data releases
	The data released often have little to do with accountability	Incremental openness based on value propositions
Supply-side of transparency	The state is complex and partly privatized	Targeted transparency (section 4.1)

3.2.1. Demand-side for transparency

In reality, the public is inhomogeneous and ‘divided.’ Furthermore, it is rather the ‘organized’ part of the public that awaits disclosure.

The ‘divided’ public. Several studies have shown that use of online data depends on individual educational, gender, income and other background characteristics of the public (Stalder 2011). The heaviest users are the youngest, wealthiest, and those with the highest educational degrees (Fenster 2012). Online participation favors highly committed minorities and special interests (Starr 2010). OGD initiatives offer advantages to ‘info-mediaries’ who develop mash-ups and applications and often (but not always) ‘translate’ the raw data for the general public (Shkabatur 2012b). Thus, the technology of choice for OGD initiatives that aim for social accountability in developing and low income countries should be first and foremost mobile telephony, the communication technology that is most widely available (Shkabatur 2012a; Fung et al. 2011; Avila et al. 2010). With 5.9 billion mobile-cellular subscriptions, global penetration reached 79% in the developing world; only 20% have Internet access (ITU 2011).

The organized and technology-savvy public. Since the general public is an unlikely user of raw open data, civil society ‘info-mediaries’ – CSOs, activists, journalists, civic hackers – that would use, adapt, and manipulate the data are the bridge between OGD and its end users. Harnessing the capacity of civic hackers is an important factor of success for any OGD endeavor. Networks of civic hackers have been active in ‘apps for democracy’ competitions although a lasting impact is elusive. For instance, dozens of apps for democracy were submitted to the Washington DC contest in 2008, visualizing data on criminal incidents, mapping the city's bike routes, and exposing the demographics of local schools. However, most became dysfunctional after the contest, because they were not responding to well identified needs (Shkabatur 2012b; Davies and Bawa 2012). The World Bank reports more encouraging results, when civic hacking is designed as a process and not as an isolated event (2012).

3.2.2. Interface between supply of and demand for transparency

The interface between supply and demand is problematic. The state often is unresponsive to the public's demands for transparency. The data released by an agency often has little to do with the agency's public accountability; the data is not accountability-relevant.

The state can be unresponsive to the public's demands for transparency. Even if mandated to disclose information, it is the bureaucracy's nature to accumulate and hoard information and to keep it secret (Stalder 2011). If we need a certain level of bureaucracy for a functioning democracy, there is bound to be a certain level of secrecy (Fenster 2006; Birchall 2011). Even with a FoI law in place, officials may evade providing the information, give mounts of irrelevant information, or outright refuse to provide the requested information (Hood and Heald 2006; Roberts, 2010). A possible solution would capitalize on the discretion of government agencies in OGD initiatives. It would require a dialog between agencies and civil society actors to persuade the former to release accountability-relevant data (Shkabatur 2012a).

The data released often has little to do with the agency's accountability. Released data could be statistics, or city master plans but also records of procedures, and decisions that led to the data, in order to foster the public accountability (Curtin and Meijer 2006). For instance, city master plans are constructed through political negotiations and lobbying by private industry (Raman 2012). Not only the city master plan itself, but also the records of the decision making that led to the master plan, the budgetary allocation decisions, and often, the choice of contractors accompanying the plan together would qualify as accountability-relevant data and open up the city administration to effective public scrutiny. A possible solution to the reluctance of agencies to release accountability-relevant data could be incremental openness. First pragmatic disclosures could be made to enhance e.g. service delivery, followed by civic disclosures of budgetary information for transparency (Yu and Robinson 2012). Thus, the active release of a machine-readable bus schedule would first enhance service delivery. Later, the disclosure of public contracting opportunities would enhance both economic opportunity and public integrity. ‘Separating technological from political ‘openness’ – separating the ideal of adaptable data from that of transparent politics – will yield benefits for all sides.’ (Yu and Robinson 2012, 24).

3.2.3. Supply-side of transparency

The state is not a coherent transmitter of the information it is required to disclose.

The state is complex and partly privatized. The state consists of a multiplicity of agencies with partially overlapping jurisdictions. Many state functions once carried out within government are outsourced and privatized. Private corporations with complex relations to the state are exempt of disclosure, and invoke commercial confidentiality to resist transparency claims (Hood 2006). A pragmatic solution for open government advocates could be to begin with targeted transparency laws and policies that compel corporations to disclose issue-specific information to achieve specific public objectives such as better education, a risk-free environment, etc. (Fung and Weil 2010).

4. Access to sectoral information

The information released in policies and interventions for access to sectoral information is issue-specific (e.g. pollution). The issue-specificity is what Targeted Transparency (TT) policies and Transparency and Accountability (T&A) interventions have in common. They differ in the direction of the flow. In TT policies the regulator compels private and public actors to actively release issue-specific, easy-to-understand facts, on e.g. pollution, to citizens. In T&A interventions, the citizen sensors themselves are the suppliers of issue-specific information, while the government is the (un)willing recipient of the information.

4.1. Targeted transparency

Targeted transparency (TT) policies overcome the complexity of the state through laws that oblige agencies (public and private) to actively release issue-specific information to achieve specific public objectives (Fung et al. 2004; Fung, Graham, and Weil 2007; Fung and Weil 2010). TT policies gradually evolved as unplanned policy innovations in response to crises, and perceived risks to citizens. Citizens perceive the released facts, embed them in their calculus and change their behavior in ways that further the aims of the policy. By changing their behavior citizens send signals to the originators of risk, who then adapt their behavior. Many TT policies are now mainstream policy tools in several countries. We use the example of management of physical risks in The Netherlands, to outline the demand-supply theory of TT policies (see Table 3) and to discuss their actual functioning (Meijer 2005).

Table 3. Supply-demand dynamics of Targeted Transparency (TT) policies.

	Supply-demand dynamics (according to TT theory)
Supply-side (state)	Crisis leads to a law for transparency of risk and active release of risk maps
Interface between supply and demand	Risk maps
	Cost-benefit calculus of citizens
	Cost-benefit calculus of originators of risk (companies)
Demand-side (public)	Public outcry triggers the policy
	Citizens duly informed, after policy is implemented; risk management improves

4.1.1. Supply side of transparency

In May 2000, a fireworks storage facility exploded. It was located in the midst of a densely populated neighborhood in the city of Enschede. The result was loss of life, hundreds of injured civilians, and the destruction of an entire neighborhood. The public outcry led to a law in 2004 obliging Dutch provinces (and later cities) to actively release risk information and make the risk transparent to the public on the internet.

4.1.2. Interface between supply and demand

The active release of risk maps by government is the centerpiece of the interface. The maps display sources of risks (factories, gas pipes, storage facilities of toxic materials), vulnerable targets (schools, hospitals, shopping centers), and the results of official inspection of hazardous facilities in some cases. The perusal of the maps is supposed to trigger a cost-benefit calculus from citizens, as well as from the originators of risks. The citizens' calculus can result to either pressure (voice) to the originators of risk (owners of hazardous facilities) or to a move to a different location (exit). The signals from the citizens' changed behavior are supposed to enter the cost-benefit calculus of the originators of risks, who then improve their risk management.

4.1.3. Demand side of transparency

The transparency cycle starts with public outcry and ends with a better informed citizenry and improved risk management.

In reality, risk management improved but not because the citizens engaged in a cost-benefit calculus or put any pressure on governments and the originators of risk. They did not. Meijer's (2005) study revealed that the policy worked because the companies themselves increased their compliance as soon as the risk maps became publicly available. He found that the majority of risk map users were not the public at large, but organized professionals, real estate agents or civil servants responsible for spatial planning. Thus the main actors on the demand side were professionals and organized users and not a nebulous public at large. The political economy of TT policies offers many insights on the actual workings of such policies, but the most salient feature is the notion of ‘organized users’. The extent to which organized users benefit or not from targeted transparency is a significant determinant of the policy's survival prospects (Wilson 1980; Fung, Graham, and Weil 2007).

4.2. Transparency & accountability (T&A) interventions

While targeted transparency laws are enacted in response to crises, the drivers for horizontal transparency, especially in middle and low-income countries, are governance failures. The shift toward horizontal transparency and social accountability (World Bank 2004) has spawned several T&A interventions (e.g. Avila et al. 2010; Fung, Russon Gilman, and Shkabatur 2010; McGee and Gaventa 2010). We discuss below two of the most promising.

4.2.1. Uchaguzi and the constitutional referendum 2010 in Kenya

The first example (fast T&A intervention) refers to elections, a government function that unfolds over a very short period of time every few years, and needs to be subjected on the spot to public scrutiny (Avila et al. 2010). For the Kenyan constitutional referendum in 2010, the Uchaguzi (‘election’ in Kiswahili) organizers used an advanced version of Ushahidi (‘witness’ in Kiswahili), a kind of Ushahidi 2.0, as a tool to crowd-source reports on electoral fraud. They created key alliances, with a NGO able to train 500 official electoral monitors on how to text reports on electoral incidents to the platform, and with the official regulatory election body in Kenya, the Interim Independent Election Commission (IIEC). The Interim Independent Electoral Commission of Kenya (IIEC) replaced the disbanded Electoral Commission of Kenya that was widely blamed for the election violence after the Kenyan general election in 2007. Electoral issues were reported to Ushahidi, either by the official electoral monitors or by members of the public, but also to the election regulator, who took immediate action in several instances, after verification. Thus, fast T&A interventions like Uchaguzi can produce quickly relevant information that can be used to exercise or demand accountability on the spot, by mobilizing a multi-stakeholder alliance of trained reporters, other NGOs, the election regulator and the government itself.

4.2.2. Twaweza and public services in Tanzania

The second example (slow T&A intervention) refers to failed public services, a government function that unfolds over a long time period. Twaweza (‘we can make it happen’ in Kiswahili), a Tanzanian NGO and member of the steering committee of the global Open Government Partnership (OGP) acts as a watchdog for three specific policy domains: education, water and public health services. Twaweza conducts extensive in-house research and weekly surveys of representative panels of trained citizens via mobile phone on their perceptions of public services. Twaweza synthesizes the results in plain language reports and graphs and mobilizes activists and local mainstream media (newspapers and radio stations) to widely disseminate policy failures. Their media coverage is substantial and often the regulator follows up, as the head of Twaweza, reported in the OGP Brasilia Conference in 2012 (Rajani 2012). Water kiosks are repaired, and exorbitant water prices are lowered. Twaweza's modus operandi is ‘quick, simple, constant bottom-up feedback informing policy.’

T&A interventions, especially of the slow kind, have a crucial characteristic that sets them apart from the collaborative production of geographic information (e.g. OpenStreetMap). Geographic information volunteered by citizens needs to conform to scientific standards to pass the quality test (Haklay 2012). The same rule applies to data collected by citizen sensors in T&A interventions – verification and sound analysis of data sourced from respondents are indispensable. The difference is this. Data in slow T&A interventions, e.g. on the shortage of drugs in clinics, or on the state of literacy of children in schools, need to be accompanied by data on officially earmarked financial resources and actual expenditures per clinic or school. Only when the deviation between funds that actually reached the clinics or schools and the officially earmarked funds is dramatic, can the collective data set – disaggregated data on drug shortages per clinic or state of literacy per school together with the respective deviations in funds – trigger corrective action from government (Rajani 2012). Thus, we can argue that the rigor of OpenStreetMap rests on whether the ‘standard deviation’ of the data is low and comparable to authoritative geographic data. The potential of success of T&A interventions seems to rest on a combination of high quality data and a ‘dramatic deviation’ between earmarked and actual official disbursements of funds. Data with low standard deviation on the state of public services if unaccompanied by data on ‘dramatic deviation’ of funds are not likely to make a ripple with the target organization. Only this combination seems to result in accountability-relevant data.

4.3. Collaborative transparency & accountability (T&A) interventions

Collaborative T&A interventions involve genuine collaboration between the public, government and an assemblage of info-mediaries. Collaborative T&A interventions are ‘efforts that intend to organize and publish public data that the government itself needs to better fulfill its regular duties.’ (Avila et al. 2010, 21). The government is not the (un)willing recipient or target of the intervention but an active participant, convener or facilitator. This new type of intervention employs mobile tools and the internet ‘a user orientation and a government facilitating role in order to create adaptable, real time, customized information that reduces risks and public service flaws’ (Fung Graham, and Weil 2007). The only genuinely collaborative T&A intervention we are aware of in East Africa is Majivoice (‘maji’ means water in Kiswahili). It has not been implemented yet, but exhibits design features salient to collaborative T&A approaches. Majivoice aims to improve efficiency, accountability, responsiveness and transparency of urban water service providers in Kenya. Here we briefly outline its characteristics, sourced from the World Bank's Water Hackathon report, May 2012.

MajiVoice and urban water services in Kenya (World Bank 2012). Majivoice was the winning application at the Nairobi Water Hackathon, the Nairobi part of a global hacking event that took place simultaneously in several cities around the world. The mobilization of civic hackers was at the heart of the World Bank global Water Hackathon in 2011, a unique gathering of software developers, governments and civil society with the aim of increasing awareness of water sector challenges facing developing countries.

The hacking challenge addressed by Majivoice was issued by Kenya's Water Services Regulatory Board (WASREB), the overseer of Kenya's Water Act 2002. It was the upshot of WASREB's prior experiences with a pilot network of community-based watchdogs, the Water Action Groups (WAGs). The WAG pilot demonstrated the value of citizen engagement and unveiled major weaknesses in the resolution of citizens' complaints. It also made clear that the amount of feedback generated by WAGs in four cities alone could only be scaled up with a smarter means of handling the bulky data. World Bank staff collaborated with the regulator in the formulation of the challenge. WASREB sent a representative to the Nairobi Hackathon to explain the challenge to the civic hackers. Only Majivoice could include all features spelled out in the challenge and won it. The Majivoice team received sponsorship to develop the system to maturity and obtained WASREB's endorsement to develop the initial prototype. The development of the platform is currently ongoing. The MajiVoice platform allows two-way communications between citizens and water providers. Citizens use a mobile phone or website to share their concerns with water providers about service delivery and receive timely feedback on how those issues are being addressed. Private water utilities can broadcast SMS messages to consumers on changes in rationing schedules, on times, dates and venues of public consultations and events, thereby strengthening their public relations and linkages with the client.

The Majivoice example provides a glimpse of how the characteristics of fast and slow T&A interventions can converge into genuine collaborative T&A interventions in the future. The ad-hoc multi-stakeholder alliances of NGOs, reporting citizens and regulators built around a single electoral event may become institutionalized in routine day-to-day service delivery processes, while additional actors enter the alliance, such as civic hackers, organized citizen groups (WAGs) and global institutions (World Bank), to quickly co-produce genuine accountability-relevant data. Collaborative T&A interventions may be the empirical nexus where the two components of open government – transparency and participation – meet.

5. Discussion

We moved from established scientific literature to describe FoI, TT policies, and OGD, to concrete examples to describe promising T&A interventions. All transparency types and variants can coexist and even co-evolve. OGD initiatives are likely to flourish in political environments with FoI laws in place. Similarly, OGD initiatives may be conducive to T&A transparency interventions, and may enable collaborative approaches in Tanzania and Kenya faster than elsewhere.

5.1. Shifts in transparency policies

We can distinguish a number of shifts in the logic of transparency policies over time. The first is from mandatory, passive release of general information (FoI) to active, but discretionary transparency (OGD). The second shift is toward regulating citizen behavior by releasing issue-specific information (TT). The onus is on the public to embed the issue-specific information in their decision making and to further public priorities, much like the onus is on the technology-savvy public to hack and mash OGD data. A third shift is from general release to sectoral (issue-specific) release of information. This is present in TT policies as well as in transparency and accountability (T&A) interventions. In the latter, though, the supply-side of issue-specific information is either the organized public itself (T&A interventions), or multi-stakeholder alliances between the state, civil society and info-mediaries, in the future. The fourth shift entails the widespread use of mobile and web2.0 technologies that in the eyes of openness advocates will allow the ‘crowdsourcing of government transparency’ and is particularly salient in OGD and horizontal transparency interventions (Brito 2007).

5.2. Drivers, characteristics and supply–demand dynamics

FoI laws arise and are sustained via civil society advocacy, much like T&A in initiatives. Crises are typical drivers for TT policies, while chronic crises (in other words, governance failures) drive horizontal transparency interventions, whether grass-roots led (Uchaguzi, Twaweza) or collaborative (Majivoice). Thus drivers have changed little or are being recycled over time. There are similarities in the characteristics as well. A free press is indispensable for a FoI law to function, but also for amplifying, together with other local mainstream media, the advocacy work of NGOs. Organized and professional users are equally important for OGD, TT and horizontal transparency interventions.

Two elements at the interface between supply and demand for T&A interventions seem to be crucial (Table 4). The first is accountability-relevant data. While incremental openness – first service delivery data, then budgetary disclosures – for OGD initiatives maybe a pragmatic solution for developed nations (Yu and Robinson 2012), T&A interventions require a combination of high quality data with a dramatic deviation between earmarked and actual official disbursements. Separating technological from political ‘openness’, separating the ideal of adaptable data from that of transparent politics, is not likely to work for T&A interventions. The second element is an extreme public – an alliance of NGOs, citizen sensors, professional users, local media, civic hackers and increasingly regulators – and not just individual citizen sensors.

Table 4. Elements for the design of Transparency & Accountability (T&A) interventions.

	Elements for the design of T&A interventions
Demand-side	Citizen sensors; mobile geoweb as technology of choice for data collection and visualization.
	Mobilization strategies of ‘info-mediaries’ (activists, journalists, civic hackers)
Interface between supply and demand	Accountability-relevant data & dramatic deviation
	Extreme publics (alliances of NGOs, citizen sensors, professional users, local media, and increasingly, regulators)
Supply-side	Level of democratization (FoI and/or OGD as proxies) and free mainstream media (press and local radio)

6. Conclusion

We started by asking whether citizen sensors creating and using VGI can radically increase the transparency of the state, or whether changes brought about by T&A interventions are more likely to be incremental. The analysis of the literature, scientific and gray, as well as our own fieldwork lead us to the conclusion that the effect will be rather incremental. It will be mediated by existing organizations and professional users who populate the space between the state and citizen sensors. T&A interventions are more likely to increase transparency if they amplify the efforts of traditional and new intermediaries, and not by by-passing or undermining them (Fung, Russon Gilman, and Shkabatur 2012). Designers of successful T&A interventions, especially in Kenya and Tanzania, whose governments are now embarking on OGD initiatives and designing FoI laws, will need to combine propositions from Table 2 with the new elements in Table 1. Extreme publics, mobilization of civic hackers, the level of democratization and freedom of press, accountability-relevant data are inseparable from the mobile geoweb, and individual citizen sensors creating and using VGI. The optimal matching of demand and supply side will probably determine the quality of a transparency regime in a specific context (Neuman and Calland 2007).

What are then the lessons for students and designers of T&A interventions? There are no definitive or universal lessons, but sensible, context-specific ways forward (Elwood, Goodchild, and Sui 2012). In the next few years of our research program in East Africa, we will ask which kinds of civil society organizations in Tanzania and Kenya mobilize and recruit citizens and/or professional workers to regularly report on the quality of public services. How can existing networks of formal and informal actors, mobilizing organizations, and traditional media benefit or lose from the emergence of the mobile geoweb? How can citizens benefit or lose by privileging large scale publicity of their grievances over face-to-face interactions with trusted intermediaries? In short, we will favor an incrementalist approach and ask how existing voices can be amplified and how existing citizen–state relations can be thickened with the mobile geoweb and citizen sensing. This way we can best further our understanding of evolving society-beneficial technologies, current and newly emerging, and advance the mission of the digital earth society.

Acknowledgments

We gratefully acknowledge the financial support of the research program, Sensors, Empowerment and Accountability (SEMA) in Tanzania, funded by the Netherlands Organization for Scientific Research – Science for Global Development (NWO-Wotro). We thank the anonymous reviewers for incisive comments that led to a revision of the first version of the paper.