Article Preview
TopIntroduction
With the ever growing pervasiveness of networking, massively distributed decision situations are becoming more common phenomena and their computer based supports are easier to imagine. While these situations have not been well articulated – they are already feasible and the corresponding decision support technology is inevitably bound to become a significant issue in the realm of information systems (IS).
Actual examples already exist. Consider, for instance, the personal video services. Such information and Communication Technology (ICT)-based services support the massively distributed decision process of video consumption, that is, how viewers decide on which content to watch. The design of these services is already a critical issue for technology vendors who develop application software for video pipelines and the corresponding set-top boxes (“Orca Interactive,” 2007; NeGeV, http://www.negev-initiative.org). Increasingly, set-top boxes are supporting TV viewers as they try to decide on content to watch – support that may range from the presently common access to information on the current TV programs or what is available on other channels, all the way to a much more complex decision support in the form of recommendation mechanisms. The latter may take into account viewing history, collaborative filtering mechanisms and even the viewer's own social network recommendation, where a friend-viewer votes (“likes”) for a specific content. Yet this rich set of potentially useful decision support applications/services is expected to run on necessarily thin end-point processors, forcing vendors to consider how to distribute these decision support mechanisms among the set-top box clients and a range of associated servers. Specifically, they will have to decide which parts of the data and recommendation models should be located at the frontend and which parts should be located at the backend. This is a crucial issue as – among other things – it directly affects system response time, which in turn has a direct effect on users' satisfaction (Tolia et al., 2006).
The above – already actual – example helps to concretize what we have labeled here as Massively Distributed Decision Support System (MDDSS) – systems in which large number of users engage together – and possibly collaborate – in an online decision process. Other online decision processes could be framed in a similar fashion: e-commerce marketplaces such as e-Bay, where buyers decide whether and what to purchase; handheld navigation systems, where users decide on routes and receive traffic updates as well as notifications from other users; election systems, where voters collectively decide on their preferred candidates; and Massively Multiplayer Online Games (MMOG) such as World of Warcraft, and Maple Story, in which large number of players cooperate and affect each other as they decide on their next move towards the goal of the game. The Internet age brings with it more and more instances of MDDSSs – massively distributed online decision process. McRoberts et al. (2010) assess the Internet's influence on specific decisions – from purchasing to politics, healthcare to finance. It shows for example that “users often incorporate the medium [internet] into their decision process” and that “Internet users report that online resources not only allow them to quickly and easily compare options, but also to seek out expert and peer advice that enables them to act with greater confidence.”
While retaining the fundamental and by now familiar properties of DSS, MDDSS may be more applicable today outside the boundaries of a single formal organization as there are only few examples of organizational DSSs with very large numbers of users.
The (possibly – very) large number of users raises the challenge of system performance (Laudon & Traver, 2008; Ginzburg et al., 2011). High traffic rates have also been known to cause system crashes (e.g., Charette, 2008; “High demand crashes,” 2010). System architects and designers are – and increasingly will be – called to design these systems so that they overcome performance bottlenecks and cope with the increasing workload demands in the most cost-effective way. Specifically – how to configure massively distributed decision support functionality.