3.2 My research approach

When I started to work on this thesis I did not have any clear understanding of neither FLOSS nor HISP. I fact, I had only limited knowledge about action research and IS theory. The beauty of action research is that the cyclical nature allows you to start out with a “fuzzy” research question and a limited understanding of the problem domain. During the advancement of my research I could constantly refine my questions and methods, and thereby give better answers.

My motivation for this thesis was to increase my understanding of FLOSS and developing countries, and how FLOSS can benefit developing countries. To do this I have read extensively about FLOSS, HISP, developing countries and Ethiopia. Most importantly, however, I have participated in the development of a prototype HIS in Tigray and the development of the next major version of DHIS.

First I will give a short description of the two cases I have participated in and state my position and biases relating to this cases. Then I will give an outline of the concrete methods I have been using during my research. I have framed my research within structuration theory so I will give an description of how I will use this as a framework for my research. Last I have to specify what I understands to be the limitations in how I have conducted my research.

3.2.1 Working in the Tigray HISP team

This case study was conducted from July until the end of October 2004. The case study took place in Ethiopia. In Ethiopia I worked together with a team consisting of me and three others. The three others where Ethiopian nationals, as a Norwegian national I was the only foreigner on the team. Of the members in our team, one was a PhD student and the other three of us were master students. The research setting was in the primary health care system of Tigray. Ethiopia is a very ethnically diverse country, so all of us on the team was more or less foreign to the region of Tigray. Most likely I was more foreign than the others in our team, I do not speak Amharic and I am a white man from a developed country.

Our position in the relationship with the Tigray health bureau and the two districts we worked in was clearly an outsider team collaborating with insiders. The collaboration took shape through the forming of a team to decide on which data elements should be included in DHIS, and on which reports the system should be able to produce. During the training sessions we got constructive suggestions on how to improve the prototype we made. An important part of our purpose in Tigray was to develop an evolutionary prototype based on DHIS for the primary health system in Tigray. Our main purpose was to help health workers and management to collect more reliable and useful data. A purpose the stakeholders naturally shared, at least in principle. We did not have time to develop a close relationship with the stakeholders, we spent in the vicinity of four months on this case, only two of them were spent in Tigray.

In the team I was an insider, but because I was the only one in the team not being an Ethiopian national I was in an other sense an outsider too. All on the team spoke adequate English, so language for internal communication was not a significant problem, except for some boring dinner conversations conducted in Amharic. I don’t speak Amharic so all conversations conducted in Amharic are invariably boring. Language was more of a problem for my relations with the stakeholders, information will invariably get lost when a discussion is translated or retold after the discussion is finished.

Our case study in Tigray was only one of several AR case studies conducted in Ethiopia. Ethiopia is a node in the HISP network and there was activity in five regions when I was there, with different teams in each region. I will explain more about the HISP network in chapter 6.

3.2.2 Participating in the development of DHIS 2

This case study have been conducted part time from the start of 2005 until the start of 2006. I do not regard my participation in this project as finished yet, but because I had to start writing my master thesis I stopped to be active in this project from early 2006 and on. Hopefully I will be able to participate more when I am finished. This project is more in line with how FLOSS projects of some size are driven.

For reasons given in chapter 10 it was decided to do a total reimplementation of DHIS. The early planning for this huge undertaking had already started some time before I signed up to write my master thesis as part of the HISP network. As I am a computer scientist who likes to program I found it appealing to work on this reimplementation. After coming back from Ethiopia I therefore started to participate in this project. In the first semester of 2005 I participated in this project through a graduate course, after that time I still spent some time coding but did not actively participate in the virtual community of the project.

This case study is more of an insider or practitioner study. During the course of the case study I contributed somewhat to other projects, in this setting I was more of an outsider. However, the perspective in this case study is mostly that of a practitioner. My purpose for doing this case study was to learn more about FLOSS through action and participation. I have to state that I was sympathetic to FLOSS from the outset of my research. To hand out code for free (as in freedom) use I do regard as a sympathetic thing to do. I had also previously been exposed to FLOSS software, like Linux, which I found fun, and sometime frustrating, to work with. This frustrations forced me to learn more about the inner working of a computer, which is a good thing for a computer scientist to learn.

3.2.3 Methods for data collection

In AR it is best to be flexible in the choice of methods used for data collection, question that demands an answer constantly appear when going through the research cycle. Some of this question are best addressed with the use of qualitative data, others can be better addressed with quantitative data. Most of the data collected are of the qualitative sort, but for the investigation of FLOSS in Ethiopia and for the investigation of the DHIS 2 project, I have collected some limited quantitative data. To collect quantitative data is a time consuming process, luckily I can use quantitative data others have collected by using existing statistics.

Observation, participation and training

The role of an action researcher is not that of an objective observer, but an action researcher still observes. Observations, and simply talking to stakeholders in the day to day work of training and making the prototype, or as I did in the DHIS 2 case, sending e-mails, is an important source of information.

In the Tigray case, the negotiations we had with the regional health bureau to set up a client-system infrastructure for our research was an important source of information. In the meetings we had with different authority persons in the bureau I took some scrappy notes as time and translation allowed. By combining this notes with my memory and the memory of other team members, I could write it up in my diary. The meetings could have been recorded, but the stakeholders did not approve of it.

We conducted training sessions at the bureau and in the two health districts. The training session at the bureau lasted over one week, in the district we had a two day training session at each site. I participated in the training session at the bureau and in one of the districts. Through this sessions I got valuable feedback. The decisions on how to conduct this training sessions was done in a partly participative manner, but the unit leader had a bigger say than the computer clerks.

In the DHIS 2 case, most of the communication were done virtually which means that the communication is more traceable. I have saved the e-mail I have sent to the different projects I have interacted with. E-mails sent to e-mail list are stored in e-mail archives available thought the public Internet. Some information are made available through the DHIS 2 Wiki. I also logged conversations I had through instant messaging.

Literature, electronic archives and statistics

My primary source of secondhand information, was information accessible through the Internet. For information relating to FLOSS this is a natural place to look. For books that were not electronically accessible I sometimes used the university library, at the University of Oslo.

For information relating to the Tigray case I used articles I found searching the Internet, international news media and statistics available from different organisation working with Ethiopia or developing countries in general. In Table 3.3 I have made a list of the most important sources of information.

Table 3.3: Internet Information Sources for the Tigray case

For secondhand information relating to FLOSS I have used sources too numerous to list here. Important references are in the regular reference section. It is common in FLOSS projects to have publicly available e-mail archives, discussion forums, Wikis, issue tracking systems and other technologies used to facilitate cooperation. Most of the information I have used I got from regular books and articles about FLOSS. If I had time to go through some more AR research cycles I could have investigated more archived information.

Narrative - Recording a diary

During my stay in Ethiopia I regularly kept a diary. In the chaotic AR research setting of Tigray this was my most important method to record my experiences and to remember names and conversations. To get the most reliable information this should be done at the end of every day, or sooner if you need to write up some scarp notes while it is still fresh in memory. If the day was rather eventless there is limited need to write an entry in the diary, but it is a good habit to have. I have to admit, however, that as our case study progressed and more demands was laid on me, my diary suffered. In the beginning I wrote frequent diary entries, but the frequency became less as the case study progressed.

Automated collection of quantitative data

On all the computers that makes up the Internet there is stored massive amount of data. A lot of this data are of the qualitative sort, but it is also possible to collect quantitative data for analysis. Unlike surveys and questionnaires this do not need to involve lengthy processes of interviewing to gather the data. The data is already “out there” in a form accessible to a computer program. By simply using a program this data can be collected. If there is no program available fit to collect the needed data, it is possible to make one which do. Two papers I have read (Lancasshire 2001) and (Mockus et al. 2002) did this. Both made Perl script to collect and process data. I have done this in my research on three occasions. On one occasion I made a python script to search for Ethiopian web servers on Google and on the second occasion I made a simple Perl script to process Linux credit files. On the third occasion I made a Perl script to gather information from the Subversion repository used in the DHIS 2 development.

To gather information about the topology of the Internet and information about the hosts connected to the Internet there exist network analysis tools. I have used two such tools traceroute and nmap. traceroute is a program useful for discovering the topology of the Internet. I used this program to find the international Internet connection of Ethiopia. nmap is a network analysis tool. This tool should be used with caution, it can potentially be ethically questionable because it is a tool that can be used to gather information that can be used to crack into a private networks. This tools is useful for discovering hosts, and to get information about them. I used this tool to find web servers running within the Ethiopian IP range.

Why I didn’t do formal interviews

When I first came to Ethiopia I had a plan to do some formal interviews, of the semi-structured or the structured kind. This is a common method used to collect qualitative data. I soon discovered that I did not have sufficient knowledge of FLOSS to ask useful questions. I was a foreigner in Ethiopia so I had no idea of who I could ask. Ideally I should have prepared more before I went to Ethiopia, but there was a need for people in Ethiopia quite soon after I had volunteered to go to Ethiopia. Part of the cyclic nature of AR is the constant refinement of research questions and methods based on experience, so if I could go to Ethiopia now I would be able to ask useful questions. I made one semi-structured interview with one of the managers of the DHIS 2 project, Knut Staring. This was done mostly to verify my interpretations of the DHIS 2 project.

3.2.4 How I will use ST

I have placed my research within the theoretical framework of ST (described in section 2.1.2). Here I will describe how I will use this framework on a practical level.

Social systems in Gidden’s definition are social practices replicated through time and space. This leads me into thinking that history is important to understand why the social systems are the way they are today. For this reason I will give relative (to the length of this thesis) lengthy historical accounts for especially FLOSS, but also for Ethiopia. By observing the social systems over an extended period of time it is possible get a better understanding of the social systems, and a better understanding of how and why they change.

Even if ST focuses on the replication and change of social systems over time and space I think it can be useful to give a snapshot image of the current situation in a social system. The model I have developed to give this snapshot view clearly resembles Gidden’s analytical model of the duality of structure (see Figure 2.1). In my model I seek to identify important modalities in the three dimensions contributing to the structures; signification, domination and legitimisation. In Figure 2.1 the focus is on the modalities and the images I have used are only meant to illustrate the model. In actual use I will not necessarily use images to visualise the modalities, more likely I will use textual descriptions. I will only use this model in my analysis of the broad social system of FLOSS. I did not get a sufficiently deep understanding of the primary health system in Tigray to model it in this way.


Figure 3.3: An analytical model loosely resembling the duality of structure

3.2.5 Limitations in my research approach

Doing formalistic generalisations based AR must be done with extreme care, AR do not naturally lend itself to formalistic generalisations. For naturalistic generalisations the responsibility for justifications is not mine, but the researcher using my research have to do the justification. In forming this thesis I have relied on other AR studies, which I hopefully have provided sufficient justification for. To make more formal generalisations the number of research cycles are important. In my study I have only went through one full circle in each case study. I have constantly refined my understanding of the research domain and I have made use of different methods along the way, but this has only been within the phases of the circle in Figure 3.2. The reason that I have only went through one circle can be attributed to the time constrains of a master thesis.

I should also have subjugated my research and interpretations to more peer review. Unfortunately I do not have an extensive network of friends with knowledge about FLOSS, Ethiopia and developing countries. I have also been hard pressed for time so I have not prioritised searching for people willing to read through my thesis (which is not a small task). The evaluation and specifying learning phases have not been done in a participative manner.

Even if I had predominantly good relations with my team members and the staff at the research sites in Tigray, I was a foreigner. The people I worked with spoke English, but some of the meetings and a lot of the more informal conversations were conducted in Amharic. Because of language problems and because I didn’t have deep knowledge about the culture of Ethiopia and Tigray, I might have missed some important clues.

To facilitate sustainability of our effort it is best with many research sites. This is more carefully explained in section 6.2. We only had access to two pilot sites in addition to the regional health bureau. The management of the bureau gave us access to only two districts. There were also several factors limiting the value of participation. There was a high rate of staff turnover, the bureau head was changed between our first and second visit. The staff at the bureau was also hard pressed for time in our first visit, an annual report on the health in Tigray had to be finished. Fractionating between the different departments in the bureau also put its limits on cooperation.

Because I didn’t record the meetings we had with the bureau I might have missed some important information. I should also have been more rigorous in maintaining my diary and faster at writing up my scrap notes. Considering my limited experience and the chaotic conditions of the research setting I managed to keep a quite good record of my research.

My participation in the development of DHIS 2 too soon became a solo effort. I, together with an other master student, started on developing a plug-in framework for DHIS 2. The other master student participated while the graduate course we took lasted. This effort became too difficult for the state of the technology at this time, and because we were not seasoned Java programmer and had limited time. We had a too grand vision for our plug-in framework. After the end of the graduate course I was on my own, and I was swamped into the coding of the framework. This limited my participation in the project as a whole. My effort became isolated from the rest of the project. I should have been more chatty about what I was doing.