Bootstrapping networks, communities and infrastructures. On the evolution of ICT solutions in health care

Transcription

1 Bootstrapping networks, communities and infrastructures. On the evolution of ICT solutions in health care Ole Hanseth & Margunn Aanestad Department of Informatics, University of Oslo, Norway. Address for correspondence: Ole Hanseth Department of Informatics P.O.Box 1080, Blindern NO-0316 OSLO Norway Telephone: Telefax: Abstract The well-known concept of critical mass focuses on the number of users as a significant factor of network growth. We argue however, that we should not only consider the size of the network, but also the heterogeneity of its elements. In order to discuss heterogeneity along several dimensions, we find Granovetter s and Schelling s models of diversity in individual preferences helpful [4-5]. In addition to the heterogeneity of the individual users, we discuss heterogeneity related to use areas and situation, to technologies, etc. The interdependencies and possible conflicts between these dimensions are discussed, and we suggest bootstrapping as a concept to guide the navigation/exploitation in/of these dimensions. Keywords: Network growth, critical mass, bootstrapping, telemedicine

2 1 Introduction Telemedicine, using multimedia communication technologies, is expected to enable improvements of health care services through radically new and improved ways of collaboration and organizing within the health care sectors. We share these visions and beliefs, and want to address the challenges of introducing and deploying telemedicine technologies. Telemedicine may potentially be used within any discipline and between all kinds of organizational units in health care. This implies that the technological solutions used by one group will have to be linked to and integrated with solutions used by others. Together this makes up a multiplicity of overlapping and interconnected networks. The various technological solutions will be integrated into one common network, a (in principle) global infrastructure. Such large and comprehensive network has to be built step by step, piece by piece through a long evolutionary process. The solutions also will have to be continuously changed and improved to facilitate the learning and organizational change required to transform the health care sector from what it is today into a radically changed and improved one utilizing the potential of the technology. Such change is not possible to predict and specify today it has to be discovered along the way. Further, as the solutions grow, to improve them involves a huge number of developers as well as users. They will be largely autonomous, deciding how they will change and use their solutions independently. Shared solutions will be produced as the aggregate outcome of the actions of independent actors not an overall master plan. Managing (or rather governing) such an evolutionary process is indeed challenging, because the network will be developed and used by large communities rather than closed hierarchical organizations. The number of members in such a community is much higher than that of organizational units and they have a high degree of independence. There is no boss that has the authority, power or legitimacy of making decisions concerning the individual members. We will in the following sections address some challenges related to evolutionary design of such networks. First we will

3 look at the logic of large networks, then on how to deal with the heterogeneity and finally appropriate design strategies that can be derived from this. 2 The logic of networks 2.1 Network externalities and increasing returns The logic of networks has been analysed within the field often called network economics (See [1] for a presentation of this theory. ) Scholars working in this field have studied the development or evolution of various large-scale and networked technologies like telecommunication and other classical infrastructures. The maybe most crucial aspect of such technologies is that their value for each user increases with the total number of users that are using the technology. The value of technologies like telephones and is not primarily related to the functions they offer the users, but to the number of persons one can communicate with using the technology. The value may, like in this example, be direct, i.e. the number of possible communication partners, or indirect, through the increasing availability of services and products related to the technology (network externalities). As the number of users grows, the technology tends to get momentum and it starts growing through a self-reinforcing process [2]. The network continues to grow by itself ; the more users that have adopted it, the higher the use value of the technology is, and the more new users will adopt the technology, and so on. This fact has several important consequences for how such technologies are or can be developed. First, it is hard to get started in the sense that no user wants to be the first one. Everybody wants to wait until or to see if others are adopting the technology. Accordingly, it is most rational for each individual not to adopt a technology before a significant number of others have done so. This is true for each individual, and if this was the sole mechanism at work, no one would be the first to adopt the technology and the network would never be built. Second, successful network building efforts need to draw upon such self-reinforcing processes where the network to a large extent takes over the

4 responsibility for its own growth. Designing networks implies that one has to manage to start such a selfreinforcing process. Third, one has to avoid being trapped and locked-in by these self-reinforcing mechanism. Gateways may be a key tool in this respect [3]. This issue will, however, not be further discussed in this article. Critical mass is a concept often employed in relation to these challenges. A common strategy in line with this concept is to identify and subsidize a number of users willing to adopt the technology. If this number of users is high enough, the network will start to grow by itself. But this strategy only works where there is an agent who may do this subsidizing. This is not necessarily the case with regard to telemedicine. The critical mass model is very simple which makes it very powerful. But in the case of telemedicine networks, we need richer models. The critical mass model assumes all elements (users and developers, organizations and institutions) to be equal. Obviously, they are not. We will here be particularly concerned with the heterogeneity of the user community and how to account for this in the evolution (i.e. design and diffusion) of the network. Mark Granovetter [4] and Thomas Schelling [5] have presented rather similar models to account for some forms of heterogeneity. 2.2 Accounting for heterogeneity: Granovetter s and Schelling s models of dying seminars Granovetter [4] and Schelling [5] both point to the fact that individual preferences vary. In many cases our preferences are not static and given, but dependent upon other peoples actions. This implies that the unfolding of various processes, like adoption of technologies, depend on how individual preferences are distributed among the members of a group or community. To illustrate this Granovetter and Schelling describe wellknown phenomena like dying seminars and pedestrians behaviour when walking across a street. In the first case, a group of students or scholars agrees to start a weekly seminar on a given topic. The first meeting gathers many interested participants, the next some less, and for each

5 seminar the number of participants decreases until the few left agree that it is better closing the whole seminar. In the pedestrian case, we often see a number of people on the sidewalk waiting for green light. If there are no cars, after a while one person may walk on red, then a little later a second do the same, then the third one, and finally the whole flock may be with a few exceptions is walking across the street. In cases like these small changes in the distribution of preferences can have tremendous effects on the outcome. To illustrate the effect of variations and changes of preferences on the outcome of processes, Granovetter [4] constructed a case: Let us assume that twenty people are waiting for green light to cross a street. Further, lets assume that among these there is one that crosses the street on red independent of any other, provided there are no cars. There is also one that walk across the street if there is at least one other doing so, one that walk across if there are at least two, and so on, up to the last one which walks across if at least 19 others do so. In this case, all individuals will cross the street. However, if we make a minimal change in the individual preferences so that one of this crowd, no. x, requires that x rather than x-1 persons must cross the street before she follows herself, the process will stop at this position. x-1 people will walk, no. x doesn t walk and consequently the rest of the group will wait for green light. If this individual (no. x) is the first one, all will be waiting. The implication of this model is that rather than buying (subsidising) users until critical mass is reached, one has to identify the users being willing to adopt the technology first, then those willing to adopt to it as second, and so one. But this is not as simple as it seems, since users preferences regarding technologies are not given, but may change. They depend on many factors, among them the design of the specific technologies. Accordingly, building a large network requires partly identifying user preferences and then sorting the users according to this, and partly shaping user preferences. The latter may be most important. User preferences may be shaped in different ways as illustrated below. Further, our overall preferences regarding a specific issue is the result of our preferences regarding many different aspects of the issue.

6 3 The multi-dimensional heterogeneity of telemedicine networks We will now turn to a discussion of the user preferences relevant for the development, adoption, and use of telemedicine solutions. Some of the preferences are mere personal, others relate to the tasks the persons are engaged in and to their working context. Also the technology and the larger (techno-institutional) context are important. 3.1 User preferences Motivation Among medical doctors (and other personnel categories within the health care sector) the attitudes towards technology vary a lot (just like within any larger group). Some, for instance, see great possibilities for improving existing and developing new medical technologies that significantly can improve the quality of the total health care services delivered to patients. Others are more concerned about the importance of the human dimension of health care and are worried that this aspect disappears in the uncritical fascination with complex new technologies. To succeed in the enrolment of the first users of a telemedicine network, highly motivated users need to be identified users who believe telemedicine technology may be designed and used in ways adding important qualities to the care processes they are involved in. Later on less motivated users may be enrolled, as: the network grows and the use value the technology increases (e.g. more users to communicate with); the technology is improved and making it easier for users to overcome barriers; the technology and the procedures it supports are improved due to use experience as well as the sceptics criticism; the sceptics become convinced about the positive contribution of the technology.

7 3.1.2 Knowledge about the technology The costs of adopting a technology depend about the adopters knowledge about it. The more knowledge about the technology, the easier it is to adopt, since the user have some idea of what it may be used for. Accordingly, the most knowledgeable users should be enrolled first. The knowledge of potential users will grow as the technology is getting closer to and used more extensively in their working environment. 3.2 Use areas and situation Personal differences between users, like motivation and knowledge, may certainly influence the adoption of telemedicine. However, most factors determining users willingness to adopt a technology like telemedicine infrastructures are not personal aspects of the users, but rather aspects of their work situation. We will here discuss some Available resources Adopting a new technology requires some resources. First of all the resources to obtain it is necessary (either money to pay for it, or e.g. contacts and industry partners that will lend it for demonstrations and tests). Secondly, the chance to experiment and learn encompass the time available, as well as a certain level of knowledge and human resources (e.g. technical skills) Types of use In some cases, the value of a network technology for a particular user is higher if there are few users instead of many. Some examples could be that a small network is simpler to use in an ad hoc manner as it doesn t require extensive coordination mechanisms, or that the access to limited resources is better. Accordingly, successful network building depends on identifying potential users and use areas where the productivity or quality of the work can be improved by communicating more efficiently with just a small number of users. To expand into areas where a large size is necessary, may then be a later step.

8 In general, the use value of broadcasting services (one-tomany) is independent of the number of users. This presupposes that an information provider is willing to invest the required resources to broadcast information even before a large number of users are connected. This is not possible if the information provider needs to fund its publishing activities through incomes from the users. If the information providers and consumers are the same, reciprocity in terms of work and benefit are required (whether it is one-to-one or many-to-many). Consequently it might be harder to make such a network start growing. One strategy is, then, to link it to an existing network and in that way draw upon the use value of that one Critical or non-critical activities Some envisioned use areas for telemedicine concerns remote assistance in acute and critical cases (e.g. complications during expectedly routine surgery). In such critical situations it is most important that the persons involved can attend to the task instead of having to focus on the telemedicine technology. Accordingly, the users should start using and obtaining experience with the technology in noncritical activities and situations. A typical non-critical situation where telemedicine technology may be used is for transmission of teaching (e.g. transmissions from live surgery), or for lectures and meetings Simple or complex practices Just as there is a huge difference between critical and noncritical activities, there is a huge difference between the degree of complexity of different medical procedures. The complexity of a procedure can be measured in terms of the number of people, medical specialities, personnel categories and organizational units involved, as well as the duration and number of steps in the procedure, the amount of technical equipment involved, or the amount interaction and coordination required. To make a network start growing, it is best starting with the simplest practices Organizational changes Telemedicine may support and improve current work practices, but many of the envisioned benefits are related to

9 its assumed potential of radically changing and improving working practices and organizational structures. The health sector is very complex, and processes aiming at radical changes and improvements will require experimental learning. Accordingly they will require time, be more expensive, and their outcome will be uncertain. Accordingly, it will probably be easier to introduce telemedicine technologies in use areas where they can support and improve existing practices. An illustration to this is that it has proved to be easier to adopt telemedicine technology in radiology than in surgery. Within radiology there are established practices of sending images to more experienced and knowledgeable radiologists for advice (second opinion), while the practice of advice in surgery is mainly based on physical co-presence in the operation room. 3.3 Aspects of technology Whether a user wants to adopt a technology or not also depends of course on aspects of or related to the technology. This include issues like availability: the users need to have close contacts to designers and support personnel; simplicity: the solutions need to be especially designed to support the actual practices; no extra functions or complexity; they need to be easy to learn and easy to use; costs: the solutions should be cheap; flexibility: the solutions may be used in many different ways so that working practices may evolve and improve without changing the solution, but the solutions should also be easy to change it when necessary [6,7]; future oriented: low risk for adopters of being trapped in a development trajectory which will prove to be a blind alley. 3.4 Coordinating institutions The development of large scale networks requires some kind of coordinating and governance structures including some kind of standardization bodies. When the network is growing, the needs regarding coordinating institutions change. In the very beginning when the number of actors

10 involved is small there is hardly any such need at all. This means that the governance institutions involved need to change. If existing standardization bodies are involved in the beginning, this may add unnecessary complexity to the project and may accordingly cause failure rather then success. 4 Bootstrapping The argument in the previous section (implicitly) says that it will be easiest to start with motivated and knowledgeable users who possess the necessary resources. A use area that doesn t depend on a large network, which is low in complexity and criticality, and which doesn t require radical organisational change will be an optimal starting point. As concerns technology choices, the best would be cheap, simple and flexible technologies for which there exist a support network (e.g. local technical support competence), then move on to more complex solutions. The underlying motivation for this advice is a bootstrapping logic. We use the term bootstrapping to conceptualise how design of networks could take the issues and challenges above into account and exploit the network mechanisms/heterogeneity. Bootstrapping is in Webster s dictionary defined as to promote or develop by initiative and effort with little or no assistance, <bootstrapped herself to the top>. It may also be described as the process of making a tool by means of the tool itself. The term is used in this sense in one case that is familiar to most of us the booting (or re-booting) of our computers, which is shorthand for bootstrapping. Many of us say that we are (re-)booting our computers rather then (re-)starting them. This term refers to the mere technical details of this process. When the computer is turned on, a complex system in many steps starts, where each step starts one system which then starts the next. For instance, the software system is bootstrapped in a process starting by the hardware reading into memory a piece of software stored in (and burnt into) a PROM and then starting this software. This software then read another piece of software starting at a fixed address on the hard disk and then starts this software. If you are booting a PC running Windows, this software will allow you to enter the setup mode (if you push

11 the F2 key within a number of seconds) if you want to boot the computer in a no-standard modus (if you, for instance, have installed several different operating systems). This software will then load and start MS-DOS, which again in its turn loads and starts the Windows system. This is a bootstrapping process because the computer s software system (or OS) is loading and starting itself. The term is used in this sense within the field of programming languages implementation, where the compiler for a language is programmed in the language itself. This is done by implementing a compiler for a minimal subset of the language in another programming language, then rewrite the complier in the subset of the language implemented. This subset of the language may then be used to make a new version of the compiler implementing an extending subset of the language. Then this extended subset can be used to implement an even larger subset and so on. And every time an error is corrected, or a new version of the compiler is made for another reason, the compiler is bootstrapped by first compiling the new version with the old one, then the new version is compiled with the already compiled version of itself and the old version (both in its textual and executable form) may be thrown away. Bootstrapping is also used in other fields. In neuropsychology, for instance, it is used to analyse and explain humans language learning [8]. The concept captures the fact that we learn new languages largely by means of the languages we already know. And within our primary language, we learn new concepts and structures by means of the concepts and structures we already know. But before this kind of learning process can start, we must as small children learn what a language and a concept are and learn our first concepts. Bootstrapping as crucial for the development of all networks social as well as technical. Another illustration of the necessity and possibilities of bootstrapping social networks is Adam Michnick s successful establishment of a political opposition movement in Poland. According to Karl Weick s outline of this story [9], all polish citizens found the communist regime so powerful and so determined in not allowing the existence of any political opposition, that they found it completely irrational to try to build such a

12 movement. One exception to this rule was Adam Michnick. He was able to find a few friends who together were able to identify a space for political action that was within what the communist regime would tolerate at the same time as this space for action was large enough for achieving some results which were meaningful those involved. And when the initial group was established and the space of action was proved to help them achieve some results, it was easier to recruit more members. And the more members recruited, the larger space of action they were able to create and the more attractive it became for others to join the movement. This strategy turned out to be very successful, leading to the movement widely known as Solidarity. The development of the Internet is the best illustration of how large scale information infrastructures are bootstrapped [10]. The Internet was established in parallel with two other networks: the Open Systems community of Unix developers and users. Later on this community has been better known as the Open Source movement developing what is now called Linux. What makes the Open Source movement work has attracted significant interest among researchers within field ranging from anthropology to software engineering [11]. Some accounts of the logic of this community describe it as status driven just like scientific communities. In such communities, members present their results (whether these are scientific discoveries or software code) to the members of their communities in order to get status among these. Accordingly, one member tries to push its result on as many fellow members as possible rather then protecting them and asking for high amounts of money to reveal them. But such a status driven community can only work as such after it has reached a certain size. And the larger the network is, the larger opportunities it gives to potential new members in terms of achieving good reputation among a large number of others. But this mechanism cannot help making a network start growing from the very beginning. The network has to be bootstrapped, and as it is growing, the status driven growth can start and accelerate the further growth of the network. This need for bootstrapping such a network can very well explain the failure of Netscape s Mozilla project where they tried creating an open source movement around their web browser software.

13 The Internet, the network of Unix users, and the Open Source movement are all networks in their own right and each of them has been bootstrapped as such. However, their co-evolution illustrates another important point: networks are often tightly interconnected into networks of networks. This is also the case for telemedicine networks. Each of the preferences mentioned above constitute a network that need to be bootstrapped. But the preferences, or networks, are interdependent. They constitute a network of networks. The overall bootstrapping process is then the bootstrapping of a network of bootstrapping networks. 5 Conclusion: Design as bootstrapping Design as bootstrapping acknowledges that users are different, and that they should be sorted according to their preferences. We should first enrol those users and practices it is easiest to enrol first, then the more challenging ones, etc. This also holds at the level of networks. We should first make the simplest one start growing, and the more challenging ones later on. In practice, however, it is not always possible to follow this strategy in a plain and simple way. The different kinds of user preferences and dimensions along which the evolution of a telemedicine network and its use should move can be modified into the following (interrelated) networks: 1. Users: Users are linked together into an increasingly larger network as more users are adopting the technology. 2. Degree of change: The health care system can to a large extent be seen as a complex machine where the different cogwheels (i.e. activities) must fit together. The higher the number of users that are working together using a shared technological network, the more they can develop new working practices utilizing the opportunities offered by the technology and which are different from existing practices. 3. More critical activities: By using the technology in one area, the users are acquiring the knowledge and experienced required to start using the technology in more complex work practices.

14 4. Knowledge: The previous point also illustrated the fact that the pieces of knowledge produced along the way are also tied together into an increasingly larger and more complex network. 5. More complex technologies: As more users are adopting the technology the technological network is growing, and as new functions are added, the network of technological components is also growing. 6. More complex coordinating institutions: The institutions which are required to reach agreement on various forms of standards (or other shared solutions and practices) need to grow as the overall network grows. This will normally include a growing number (i.e. network) of (sub-)committees taking care of specific forms of standards or coordination issues. As a general rule these networks should be bootstrapped in the sequence they are listed above. This may be expressed in a condensed way in a pseudo-code like program as follows: 1. Start by design the first, simplest, cheapest solution we can imagine and which satisfy the needs of the most motivated users in their least critical and simplest practices and which may be beneficial by supporting communication and collaboration between just a few users. 2. use the technology and repeat as long as possible: enrol more users 3. if possible: explore, identify and adopt more innovative (and beneficial) ways of using the solution, go to 2 4. use the solution in more critical tasks, go to 2 5. use the solution in more complex tasks, go to 2 6. improve the solution so new tasks can be supported, go to 2 References 1. Shapiro, C. and Varian, HR.: Information Rules: A Strategic Guide to the Network Economy. Boston, Harvard Business School Press, 1999.

15 2. Hughes, T.: Networks of Power: Electrification in Western Society, Baltimore, John Hopkins University Press, Hanseth, O. Gateways as tools and actors: How the Internet won the religious war about standards in Scandinavia. Submitted for publication. 4. Granovetter, M. Threshold models of diffusion and collective behavior. Journal-of-Mathematical-Sociology. vol.9, no.3; 1983; p Schelling, T. C. Micromotives and macrobehavior. New York : W. W. Norton, Hanseth, O., and Lundberg, N. Designing Work Oriented Infrastructures. Computer Supported Cooperative Work, Vol. 10, Nos. 3-4, Hanseth, O., Monteiro, E., and Hatling, M. Developing information infrastructure: The tension between standardization and flexibility. Science, Technology and Human Values. Vol. 21 No. 4, Fall 1996, Weissenborn, J. and Hohle, B. (eds.): Approaches to Bootstrapping : Phonological, Lexical, Syntactic and Neurophysiological Aspects of Early Language Acquisition, Vol. 1. John Benjamins Pub Co, (October 2001) 9. Weick, K.: Sensemaking as an organizational dimension of global change. In D. L. Cooperrider and J. E. Dutton (eds.): Organizational Dimensions of Global Change. No Limits to Cooperation. Sage Publications, London, UK, 1999, pp Abbate, J. Inventing the Internet. MIT Press, Cambrige, Ma., USA, Bergquist, Magnus och Ljungberg, Jan (2001), "The Power of Gifts: Organizing Social Relationships in Open Source Communities" to appear in Information Systems Journal (Special Issue on Open Source Software).