Summary:

The paper presents a task distribution system – WeBuild. As physical building tasks often requires coordination among multiple workers, one of the problems often occurred is that a large chunk of time is wasted on coordination process such as distributing main task into subtasks, assigning subtasks to individuals, etc. WeBuild takes the advantage of known benefits of task management systems and interactive instructions to collocated construction and assembly coordination, mimicking the scenarios with the existence of an experienced manager. WeBuild is based on an algorithm that considers each individual worker’s skill set, availability of the tools and dependencies among subtasks to automatically assign subtask to the right worker. Such improvement is supposed to always be supported throughout the entire process. Initially, to understand the design characteristics, the authors carried out an observational study to ask group workers to assemble a Meccano Tower Bridge set and an IKEA cabinet. Their findings shed light on design goals of WeBuild. The validate WeBuild, the authors carried out an initial study to compare WeBuild and traditional paper manuals. The suggestive findings are generally promising, but with limitations due to scalability and generalizability.

Reflection:

Being involved in group projects, though not quite the same as group assembly tasks, I totally understand the motivation of this paper. The nature of group work seems to be chaotic for me. When given a task that requires group work, it is inevitable that this task needs to be divided into multiple subtasks. Then who should be the one in charge? At any start up time, everyone had their own ideas and own thoughts. It would be problematic to decide a leader and divide the task in a way the at least appeals to most workers. That’s just the start. As a result, normally a chaos arises initially when someone might be competing to get some subtasks while some subtasks are left unattended. After subtask assignment chaos is settled, another problem reveals itself to the group. What if (which is very likely) a subtask is assigned to a worker who does not have appropriate skill for it? Finally, after all these have been resolved, how to merge all subtasks back together. From this point of view, I totally understand the authors concern and is looking forward to something like WeBuild.

On the other hand, the paper has its drawbacks as well. As the authors acknowledged, the study is at best an initial attempt. This is limited by the scale of the assembly task and generalizability of its result. Another interesting thing is about the system’s reliance on manual input, such as task specification and skill et of individual users. It is not the case that everyone knows themselves clearly and might introduce error.

One of the last things is the graphs. I don’t think having an absolute time plot right next to an average percentage histogram is a good idea because the relation and interpretation of these two visualizations are totally different.

Question:

To make the work more generalizable, on exactly what scale could be proper?