Human-Centered Computing

Human-Centered Computing

Human-centered computing studies information systems in their social contexts: The effects of those systems on individuals and societies, and applies understanding of social factors in the design of new systems. In the context of Citris, there are exciting possibilities for enhancing the value of information systems to the public:

Social Studies of SISs using SISs

Societal-scale information systems can have deep implications for the health, privacy, and full participation of users in society. A distinctive aspect of our effort is the incorporation of policy, planning, and social science faculty in close collaboration with engineers. Our design process will prioritize social considerations equally with technical ones [Castells, Saxenian]. We will perform sociological, ethnographic and psychological studies to measure the success of different systems and work to make human issues central to the design process. This comprehensive approach to SIS design will extend to the curriculum, where we will introduce interdisciplinary courses focusing on the full complexity of societal-scale systems.

Another unique aspect of our effort will be multi-disciplinary study of SIS using SISes as a tool for social analysis. The ubiquity of sensors and distributed data access makes it possible to study user behavior simultaneously at large scale and fine granularity. We have already developed tools such as WebTango and Webquilt [Hearst, Landay] for analysis and visualization of web site use based in part on user access patterns. Webtango is an automated tool designed to analyse information-based web sites. Webquilt is a visualization tool directed at understanding user traffic patterns across a general web site. Vastly more information will be available from SISes. Processing and interpreting this data is a great challenge that will require close collaboration between social scientists and computer scientists [Canny, Castells]. In future, we plan to use data from SISes to study questions such as: (i) Do users from different demographic groups have equal access to the full spectrum of information resources (including mobile access)? (ii) Is language skill a significant impediment to use of information resources? (iii) What happens to communities in highly-networked environments?

As with any study it is important to get informed consent from users, and its important to preserve the privacy of study participants to the maximum extent possible, especially in distributed, heterogeneous SISes. We have developed an efficient protocol for secure aggregation of private data [Canny, Tygar]. This allows statistics to be computed from many users’ data without exposing the data itself. This technology extends electronic voting protocols, and beyond voting, its supports new forms of collective decision-making. Ordinary citizens could participate directly and securely in policy decisions when referenda or traditional voting would be impractical.

Full Participation of Citizens in the Information Society

We are concerned that all citizens have access and the ability to fully exploit SIS technologies. Within the US, three communities in particular are often left at the margins: (i) physically challenged individuals (ii) elder citizens (iii) non-native English speakers. For (i) we have been active in the development of NSL (Natural Language and Speech) interfaces [Feldman] and multimodal [Landay] interfaces that provide a variety of options to physically-impaired computer users. The UCB and International Computer Science Institute (ICSI) multi-lingual NSL efforts are at the forefront in all of the key technologies for NSL. Within multimodal interfaces, we have developed tools for pen-based, speech-based and traditional GUI interfaces, and we are developing a tool that combines several modes at once. Our researchers have extensive interactions with California companies, including Nuance, Qualcomm, Ask Jeeves, and SRI international, and our design tools are in use by many others. A Berkeley project on drawing tools for the blind uses speech I/O and allows blind users to “view” drawings as well.

For elder citizens (ii) we will begin with social studies to better understand elders’ specific problems and needs. We will apply the principles of “human values-based design” to uncover issues that most strongly affect their quality of life: Ubiquitous sensing has been used to look for changes in elder’s usually regular habits (Atkeson, Abowd) as these often indicate physical and psychological problems. Many elder citizens value belonging to a community (such as SeniorNet) and lifelong learning. UCB has had a leading role in research on telepresence [Canny] which provides a stronger sense of social presence a remote space and can facilitate such activities for citizens with limited mobility. Other more pragmatic issues for elders are weakening memory and specific cognitive limitations on learning.

Non-native speakers of English (iii) are the fastest growing segment of the US population. These students face language-related difficulties in all subjects at school that leave them greatly disadvantaged by the time they enter the workforce. Solving the educational problems is beyond the scope of this proposal. However, SISes will be of greatest benefit to those citizens who can interact with them directly, and language skills will come into play whenever users try to manage the potential deluge of information that will be available. Information visualizations [Hearst] and manipulable representations (Gifford) can transcend linguistic differences. We will build on Berkeley work that introduced the “meta-thesaurus”, a corpus-based technique that allows search of information in one language using queries in another, and on work on NSL learning by young non-native speakers [Feldman].

Social Dynamics of Software Engineering

While new verification schemes may someday improve the reliability and efficiency of programming, today software engineering is a socially-limited process. Large program development is limited by communication (> 50% of programmers’ time) and coordination costs (Brook’s Law), and complex software systems mirror the organizational structure of the teams that developed them (Conway’s Law): e.g. failures are empirically more likely at boundaries between components developed by separate teams with poor communication. Changing the intra-team communication medium can significantly affect team productivity (Sproull and Kiesler). Ongoing work on rapid prototyping tools and coordination and awareness tools for team programming [Landay] is highly applicable here. Other work [Sack, Canny] has begun an exploration of an apparent counter-example to Brook’s law: the open-source movement and the development of Linux, which can be viewed as a societal-scale “socio-technical system”. We are studying the sources of its apparent success via factors such as persistent communication (message archiving), dynamic subcommunities, value of hybrid language (code+comments) in describing design patterns, and its well-organized knowledge network of both people and documents. Lessons learned will be applied to search tools, code taxonomies, referral systems (guru finders) and coordination and awareness tools for other programming situations.