Contextual Software

这项工作旨在推进计算机科学的一个领域，称为普适计算（ubicomp），其目的是使移动计算机，如移动电话，为使用它们的人做新的和有用的事情，并且与每个用户的环境非常契合，以至于他或她几乎不需要考虑使用它们。例如，经常参观博物馆的游客在参观一个新城市时，可能会使用手机上的当地地图。当他经过一个新博物馆时，他的地图可能会显示他的手机已经找到了显示博物馆当前展览的软件，并为他准备好了运行。然而，对于ubicomp来说，一个真正的问题是人们的兴趣和背景一直在变化。人们在他们的日常生活中，在他们选择的地方，为了他们自己的目的，以他们自己的方式使用这种技术。这使得评估变得困难，因为很难观察和理解用户不断变化的行为。反过来，这也阻碍了评估人员向程序员提供有关用户行为、问题和需求的建议。例如，游客可能会决定他现在已经受够了博物馆，跳上一辆公共汽车到城市高处的观景台观看烟花表演。理想情况下，ubicomp软件应该能够适应——或者被适应——这种新的使用环境，也就是说，它应该是“上下文软件”。一般来说，适应太不可预测，无法自动完成，而用户又太复杂，无法单独完成。游客不太可能根据自己的兴趣、观点、烟花表演和手机上已有的软件来开发新的软件。我们认为，这需要用户软件的进步，需要评估人员和程序员的可视化和工具的进步，需要将它们连接在一起，以便用户、评估人员和程序员能够更好地相互交流。我们已经迈出了一些第一步，使我们有了一个良好的开端。Domino是一个允许ubicomp用户直接在移动计算机之间共享小块软件的系统。它会推荐哪些新软件适合用户电脑上已有的软件。它的工作原理很像亚马逊的图书推荐：“买了你买过的书的人也买了……”这些”。这样的软件推荐就像亚马逊的图书推荐一样容易处理。但是，我们希望改进Domino，以便它根据更多的上下文（例如，一个人要去哪里以及他拥有的软件）给出建议。2. Replayer是一个让评估者和程序员可视化ubicomp系统如何运行以及它在过去是如何使用的系统。它“重播”系统所做的记录，与评估者的视频、音频和文本笔记同步。我们想把它扩展到可视化信息流，从移动电脑，显示用户正在做什么。我们还希望将Replayer与程序员用来编写新软件的工具集成在一起。然后，他们可以看到当前软件的细节，以及过去和当前的使用环境，因为他们可以使用新的软件来适应用户当前环境和潜在的未来环境。然后，他们可以将软件发送给那些上下文中的人，通过Domino进行推荐。想象一下，游客到达了观看烟花表演的观景台。他可能会从这个新地方的其他人那里（通过Domino）得到新软件的推荐，然后让它在他的手机上运行。这也许能解释为什么会有烟花表演，也能告诉我们去哪里看他以前去过的地方。如果没有这样的新软件，或者如果他对它有问题，可以通知评估人员和程序员。然后，他们可以使用Replayer来分析游客手机上的背景和系统，并使用适合他们的新软件进行响应。在我们的项目中，我们将演示上下文软件首次在ubicomp中显示出这样的响应性和适应性——所有这些都在用户进入评估人员/程序员团队之前未知的位置后一小时内完成。

This work aims to advance an area of computer science called ubiquitous computing (ubicomp), which aims to make mobile computers, such as mobile phones, do new and useful things for the people using them - and yet fit so well with the context of each individual user that he or she hardly needs to think about using them. For example, a tourist who often visits museums might, when visiting a new city, use a local map on his phone. When he walks past a new museum, his map might show that his phone has found software showing the museum's current exhibitions, and got it ready for him to run.However, a real problem for ubicomp is that people's interests and contexts change all the time. People use such technology in their everyday lives, in places they choose, for their own purposes and in their own ways. This makes evaluation difficult because it is hard to observe and understand users' changing behaviour. In turn, this hampers the usual way that evaluators advise programmers about users' behaviours, problems and needs. For example, the tourist might decide that he has had enough of museums for now and jump on a bus to a viewpoint high above the city, so as to view a fireworks display. Ideally, then, ubicomp software should be able to adapt - or be adapted to - such new contexts of use, i.e. it should be 'contextual software'. Generally, adaptation is too unpredictable to be done automatically and too complex for users to do alone. The tourist is unlikely to make new software tailored to his interests, the viewpoint, the fireworks display and the software already on his phone. We believe that this requires advances in users' software, in visualisations and tools for evaluators and programmers, and in connecting them all together so that user, evaluator and programmer can better interact with each other. We have made some first steps that give us a head start:1. Domino is a system that lets ubicomp users share small pieces of software directly between mobile computers. It gives recommendations as to what new software might fit with what is already on one's computer. It works rather like the Amazon book recommender: 'people who bought the books you've bought also bought... these'. Such software recommendations are as easy to handle as an Amazon book recommendation. We want to refine Domino, though, so that it bases recommendations on more of one's context, e.g. where one goes as well as the software one has. 2. Replayer is a system that lets evaluators and programmers visualise how a ubicomp system ran and how it was used in the past. It 'replays' logs of what the system did, synchronised with evaluators' video, audio and textual notes. We want to extend it to visualise information streaming in from mobile computers, showing what users are doing now. We also want to integrate Replayer with the tools programmers use to write new software. Then, they can see the details of their current software, along with past and current contexts of use, as they respond with new software to fit users' current contexts and potential future contexts. They can then send the software out to the people in those contexts, for recommendation via Domino.Imagine the tourist reaching the viewpoint for the fireworks display. He might be recommended new software from other people in this new location (via Domino), and then let it run on his phone. It might explain the fireworks display, and show where to look to see the places that he visited before. If there's no such new software, or if he has problems with it, evaluators and programmers could be informed. They could then use Replayer to analyse the context and the system on the tourist's phone, and to respond with new software that fits with them. In our project, we will demonstrate contextual software showing such responsiveness and adaptation for the first time in ubicomp - all within an hour of the user entering a location previously unknown to the evaluator/programmer team.