Summary
In this paper, the authors try to help crowdsource workers to have a big picture of the system that the small tasks assigned to them try to accomplish which can help in executing the task more efficiently to have a better contribution on achieving the big goal. The work also tries to help companies to remove the bottleneck caused by a small number of people who normally know or maintain the big picture and can cause serious risks if these individuals leave the company. The authors designed and developed a system known as Knowledge Accelerator that can be used by crowdsource worker to answer a given question and allow them to use relevant resources to help them answering the question in a big picture context without the need to a moderator. The system starts by asking the workers to choose different web pages related to the topic then extract the relevant information, then cluster the information on different categories. The system then integrate the information by drafting an article, then allow the editing of the article, and finally add supporting images or videos that are related to the article, and that way the system can help the crowdsource worker to know the big picture of the system and complete the task in a way that can achieve the big picture and goal of the system.
Reflection
I found the study mentioned in the paper to be very interesting. I agree with the authors that a lot of tasks that are done by crowdsource workers are simple and it is hard to divide complex tasks that can require the knowledge of the big picture. Knowing the big picture is very important and often known be very few people who are normally in technical leadership positions and losing them might cause serious issues.
I like the way the system is designed to provide a high-level information about the system while working on a small tasks. The pipeline and multi-stage operations used in the system to generate a cohesive article can help the workers to achieve the goal and also to know more information about the topic.
This approach can also be used when building large scale systems where many components need to be built by developers, and often these developers don’t know what the system is trying to accomplish or try to solve. Normally developers work on a specific task to add an employee table or building a web service endpoint that can receive a request and send back a response without knowing who will be using the system or what will be the impact of their tasks on the overall system. I think we can use such system to help developers to understand the big picture of the system which can help them to solve problem on a way that can make a greater impact on the big picture or big problem that the system is trying to solve.
Questions
- The system developed by the authors can help with generating articles about a specific topic, can we use the system to solve different problems?
- Can we use this approach in software development to help developers understand the big picture of the system that they are trying to build especially when building large systems?
- Can you think of a way to use a similar approach in your course project?
Hi Mohannad, to address your first question, I think potentially the proposed can generate an article to answer different problems, but it depends on how the problems are structured. One possible scenario is that the problems relate to each other as sub-problems to a topic. The proposed framework can deal with this scenario by setting a larger picture and filtering the information relevant to the sub-problems. Another scenario could be more challenging in which problems have multiple topics involved. To leverage the power of the system in that scenario probably requires a moderator (which could be individuals or automated model) to separate the problems according to their big pictures and tackle them one by one. I would imagine that the more complex the problems are, the more difficult the system can achieve the desired performance with a large number of crowd workers collaborate with the system in distributed manner.