David N. Chin

Talking to UNIX in English: an overview of UC

UC is a natural language help facility which advises users in using the UNIX operating system. Users can query UC about how to do things, command names and formats, online definitions of UNIX or general operating systems terminology, and debugging problems in using commands. UC is comprised of the following components: a language analyzer and generator, a context and memory model, an experimental common-sense planner, highly extensible knowledge bases on both the UNIX domain and the English language, a goal analysis component, and a system for acquisition of new knowledge through instruction in English. The language interface of UC is based on a “phrasal analysis” approach which integrates semantic, grammatical and other types of information. In addition, it includes capabilities for ellipsis resolution and reference disambiguation.

Empirical Evaluation of User Models and User-Adapted Systems

Empirical evaluations are needed to determine which users are helped or hindered by user-adapted interaction in user modeling systems. A review of past UMUAI articles reveals insufficient empirical evaluations, but an encouraging upward trend. Rules of thumb for experimental design, useful tests for covariates, and common threats to experimental validity are presented. Reporting standards including effect size and power are proposed.

User Modeling, Adaptation, and Personalization

Interactive technologies pervade every aspect of modern life. Web sites, mobile devices, household gadgets, automotive controls, aircraft flight decks; everywhere you look, people are interacting with technologies. This trend is set to continue as we move towards a world comprising Smart Cities built around the Internet of Things. Unfortunately, much of the rhetoric surrounding this dawning age of ubiquitous and embedded computing fails to appropriately consider the people at the centre of it. These people are embodied social agents with motivations, emotions, capabilities, capacities, proclivities and predilections. Technological imaginings around the Internet of Things are often steeped in generalities or idealised scenarios of use. Such imaginings typically forget that design is always about meeting particular peoples’ needs in particular contexts. From concept to ideation to prototype and evaluation, the design of interactive technologies and systems that are intended for people should start with some understanding of who the users will be, what tasks and experiences they are aiming for, and what the circumstances, conditions or context(s) are at play. In this talk, I will discuss a simple people-centric framework devised with my colleagues and coauthors to inform the way we think about design, the ABCS of designing interactive systems. A descriptive guide rather than a prescriptive checklist, the framework draws on basic research in ergonomics, psychology and user modeling. It is intended to focus design thinking about people as the users of interactive, computational systems. It is intended to support us as the designers of interactive technologies as we scope, draft and iterate on the design space of imagined interactive experiences. Using examples from my own work, I will illustrate how this framework has been explicitly and/or tacitly applied in the design, development and evaluation of interactive, multimedia systems. In particular, I will consider how this framework is currently being applied to rethinking the concept of personalization.

KNOME: Modeling What the User Knows in UC *

KNOME is the user modeling component of UC, a natural language consultation system for the UNIX operating system. During the course of an interactive session with a user, KNOME infers the user’s level of expertise from the dialog and maintains a model of the user’s knowledge of the UNIX domain. KNOME’s model of the user makes use of a double-stereotype system in which one set of stereotypes represents the user’s expertise and another represents the difficulty level of the information. KNOME is used in UC to help disambiguate the user’s statements, avoid telling the user something that the user already knows, take advantage of prior user knowledge in presenting new information, and detect situations where the user lacks pertinent facts or where the user has a misconception. UC also models its own knowledge of UNIX with meta-knowledge (explicit facts about the limitations of the system’s own knowledge base), which is used to help in correcting user misconceptions.

The berkeley UNIX consultant project

UC (UNIX Consultant) is an intelligent, natural language interface that allows naive users to learn about the UNIX2 operating system. UC was undertaken because the task was thought to be both a fertile domain for artificial intelligence (AI) research and a useful application of AI work in planning, reasoning, natural language processing, and knowledge representation.The current implementation of UC comprises the following components: a language analyzer, called ALANA, produces a representation of the content contained in an utterance; an inference component, called a concretion mechanism, that further refines this content; a goal analyzer, PAGAN, that hypothesizes the plans and goals under which the user is operating; an agent, called UCEgo, that decides on UCs goals and proposes plans for them; a domain planner, called KIP, that computes a plan to address the users request; an expression mechanism, UCExpress, that determines the content to be communicated to the user, and a language production mechanism, UCGen, that expresses UCs response in English.UC also contains a component, called KNOME, that builds a model of the users knowledge state with respect to UNIX. Another mechanism, UCTeacher, allows a user to add knowledge of both English vocabulary and facts about UNIX to UCs knowledge base. This is done by interacting with the user in natural language.All these aspects of UC make use of knowledge represented in a knowledge representation system called KODIAK. KODIAK is a relation-oriented system that is intended to have wide representational range and a clear semantics, while maintaining a cognitive appeal. All of UCs knowledge, ranging from its most general concepts to the content of a particular utterance, is represented in KODIAK.

User modeling in UC, the UNIX consultant

UC is a natural language computer consultant system for the UNIX operating system. The user model in UC encodes the users knowledge state and allows UC to tailor its responses to the user. The model encodes apriori knowledge in a double stereotype system that is extremely efficient. Models of individual users are updated dynamically and build on top of the users stereotype. The model deals with uncertainty in apriori information and attempts to deduce the users level during the course of a session.

Acquiring user models

Existing machine techniques for acquiring user models are characterized along five orthogonal dimensions: passive/active, user-initiated/automatic, logical/plausible, direct/indirect, and on-line/off-line. Passive techniques observe users whereas active techniques query users. User-initiated techniques require that users volunteer information; automatic techniques do not. The logical/plausible dimension measures the accuracy of derived user model data. Indirect techniques build upon data gathered by more direct methods. On-line techniques acquire user models in real-time during user interaction, while off-line techniques work after the user interaction is finished. Commonalities and differences in capabilities and features of different user model acquisition techniques are analyzed along the above dimensions, and the relationship of these techniques to similar techniques in other areas of artificial intelligence are discussed.

Acquiring User Preferences for Product Customization

Mass customization requires acquisition of customer preferences, which can be modeled with multi-attribute utility theory (MAUT). Unfortunately current methods of acquiring MAUT weights and utility functions require too many queries. In Iona, the user is first queried for absolute/preferred constraints and categorical preferences to cull the product pool. Next Iona selects queries to maximally reduce the utility uncertainty of the remaining product choices. Implemented queries include stereotype membership and contexts (the purchase situation), which give probabilistic MAUT data modeled as ranges of weights. The usefulness of a query is based on the reduction in uncertainty (smaller range) weighted by the likelihood that the user belongs to a stereotype/context based on similarity to the current user model. Querying proceeds until the usefulness of the best query is below the threshold of user impatience. Finally integer programming is used to select the best product for the user.

DECODE: a cooperative environment for reverse-engineering legacy software

While automated program understanders have had some success in partially extracting design information from source code, they are unlikely to be able to completely understand existing real-world legacy systems. To address this problem, we have been developing DECODE, an environment in which programmer and system cooperate to extract object-oriented designs from legacy systems. DECODE consists of three components: an automated program understander that extracts some initial stereotypical object-oriented design elements; a structured notebook that provides the user with a graphical view of the systems understanding and the ability to extend this understanding by linking source code fragments to object-oriented design elements; and a query processor that uses this design information to support conceptual queries about the programs code and design. This paper describes DECODE and our initial successes and failures with using it to reverse engineer several non-trivial COBOL programs.

An analysis of scripts generated in writing between users and computer consultants

The scripts generated in written interactive communications between users and a computer consultant program were investigated in a controlled experiment. The program was a simulation of UC, the UNIX Consultant, which users believed to be the actual program. An analysis of the scripts generated while solving a predefined set of problems showed the heavy use of context in forms such as ellipsis, anaphora, indirect speech acts, and grammatically incomplete sentences in over one-quarter of input clauses. Also present were grammatically ill-formed constructions and spelling errors. A comparison with a control group of users solving the same problem set with human consultants showed that the control group relied on context about twice as much as the simulation group. This suggests that people naturally use context in language and that the simulation group tried to rely less on context because they believed that they were speaking to a computer. Even so, contextual information is essential to understanding a large part of the simulation groups input.