HCI Models, Theories and Frameworks: Toward a Multidisciplinary Science
by John Carroll

Carroll, J. M. (Ed.) (2003). HCI Models, Theories and Frameworks: Toward a Multidisciplinary Science. San Francisco: Morgan Kaufmann publishers.   ISBN: 1-55860-808-7.
This collection of tutorial articles is an appropriate survey for the graduate level student. [DS]

Jump To:   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15

Ch. 1:  Introduction
Ch. 2:  Design as Applied Perception (visual senses)
Ch. 3:  Predictive vs. Descriptive Models, Fitt's Law
Ch. 4:  GOMS, KLM, Advanced / Modified GOMS
Ch. 5:  Cognitive Dimensions of Notations Framework
Ch. 6:  Users' Mental Models
Ch. 7:  Information Foraging Theory, Optimal Foraging Theory, Scatter/Gather (evolutionary)
Ch. 8:  Collaborative Technologies / Distributed Cognition
Ch. 9:  Cognitive Work Analysis (CWA)
Ch. 10:  Clark's Common Ground Theory (as used in CMC)
Ch. 11:  Activity Theory
Ch. 12:  CSCW Research (Psychological foundations)
Ch. 13:  Ethnography, Situated Action, Ethnomethodology
Ch. 14:  Computational Formalisms
Ch. 15:  Design Rationale as a Theory

Chapter 1:  Introduction

HCI lies between social and behavioral sciences; computer and information technologies.  It is the fastest growing CS field, and needs a multidisciplinary approach to keep flourishing.

HCI practitioners:  analyze and design user interfaces, integrate technology to support human activity

Initially, HCI brought cognitive science theories to focus on software development

Activity theory brought into HCI (Marxist foundations), complemented cog. sci.

While a multidisciplinary approach helps HCI theories, a tradeoff is the fragmentation of HCI (it's very difficult to make sense of the vast, diverse science of HCI; to synthesize a comprehensive, coherent methodological framework)

Goal of the book:  To 'survey' the many approaches to HCI; compare / contrast them
 - each chapter is a different approach / method


Chapter 2:  Design as Applied Perception

Much of human intelligence can be characterized as our ability to recognize patterns

Vision dominates as the primary sense:  engaging 50% of the cortex and more than 70% of all our sensory receptors

Why we care:  we have a fundamental assumption that there is such an entity as the human visual system

display design - map data into a visual form that is matched to our perceptual capabilities

In Cognition, humans share a similar muscular / skeletal structure, but have a highly adaptable neural structure, which allows for a high degree of adaptation (humans are highly adaptive machines)

Information Psychophysics - concerned with how we can see simple elementary information patterns (paths in graphs, clusters of items, correlated variables, etc.)


3 Stage Model of the Visual System:  this happens very fast - our brain's process when we see stuff

In the above system, feedback loops can modify what we see.  Higher stages can feedback to lower stages to modify what we see.  Lower stages are more robust, while higher stages involve more inference making.

In the above system, culture makes it difficult to design by visual patterning since aspects of displays owe their value to cultural factors (R, G, Y, B are different colors in every culture).  Some cultural aspects are so hard-wired that there's no way to design around them.


Chapter 3:  Motor Behavior Models for HCI

Why model motor behaviors?  Because we need to match movement limits, capabilities, and potentials of humans - to input devices and interaction techniques - on computing systems

A model is a simplification of reality that we can use to design, evaluate and provide a basis of understanding a complex behavior of complex artifacts. 

Models lie somewhere in a continuum:    analogy / metaphor  <-- ? --> mathematical equations

More on Predictive Models:

allow human performance to be analyzed analytically, avoiding time-consuming experiments

predictions are a-priori, allowing for hypothetical exploration of a design scenario

Hick-Hyman Law:  an example predictive model, predicts human response time by an equation

Fitt's Law:  a predictive model for human movement (measures accuracy and amplitude of human movement; computes an index of performance to compare efficiency of different devices / interfaces)

Keystroke-Level Model (KLM):  a predictive model for predicting time to do a task by dividing into sub-tasks:

K = Keystroking
P = Pointing
H = Homing
D = Drawing
M = plus 1 Mental operator
R = plus 1 system Response operator

TExecutive = tK + tP + tH + tD + tM tR

Other Predictive Models:  GOMS (see Ch. 4), Programmable User Model (PUM:  Young, Green, Simon 1989)

More on Descriptive Models:

Descriptive Models provide a framework / context for thinking about a problem

Example:  Key-Action Model (KAM):  divides keys on a keyboard into 3 categories:  symbol keys, modifier keys, and executive keys (simple model, organizes keys into categories)

Three-State Model of Graphical Input (Buxton):  Descriptive model, says computer pointing devices follow a state transition diagram of 3 states:  out of range, tracking, and dragging (good for analyzing different types of pointing devices [touchpad, trackpad, mouse, etc]) - resulted in redesign of touchpad's to be pressure sensitive, so drag state could be induced easily

Guiard's Model of Bimanual Skill:  studies between-hand division of labor in everyday tasks - see chart of descriptive model on p. 41

Buxton & Myers found that people prefer to use two hands when not instructed, and this resulted in lower times in task completion.  This had little insight into hand use, but was one of the first papers in HCI.

Another study (Gibson) of keyboards found that they are right-hand biased (power keys on right side), but the Desktop as a whole is left-handed bias:  lefties can cash in on a time savings, because having the mouse to the left of the keyboard allows for the right hand to use the keypad / power keys WHILE the left hand operated the mouse (don't have to stop, switch between the two)


Chapter 4:  Information Processing and Skilled Behavior

GOMS is the primary focus of this chapter (Goals, Operators, Methods, and Selection rules:  a way to describe a task and calculate the time to complete it)

GOMS applies to situations in which users will be expected to perform tasks that they already have mastered.  It does not work for tasks being done by novices that are trying out a new interface design.  The knowledge gathered by GOMS reflects what a skilled person (expert) will do in a seemingly unpredictable situation.

GOMS works for single-user, active systems where the system changes in unexpected ways or other people -participate in accomplishing the task.  It has been shown to work very well in analyzing user-paced, passive systems.

GOMS can be used quantitatively and qualitatively. 

GOMS as a task analysis technique:  stemming from conceptual frameworks on human information processing (Carroll, p. 65):

Conceptual frameworks - informally stated assumptions about the structure of human cognition

serial stage models - we process information in a serial sequence
parallel multiprocessor stage models - we process information like parallel computing

Computational Cognitive Architectures - also called unified theories of cognition - are proposals to make cognition explicit enough to run as a computer program

The Model Human Processor (Card et. al, 1983):  ties cognitive science and engineering models:  represents human cognition as a parallel computer.  The parallel processing being done:

In the diagram above, the bottom is more theoretical while the top is more practical.  GOMS is a way designed to make human information processing something that can be measured practically.

Task Analysis Techniques - these map out the system in terms of goals, operators, methods and selection rules.  There are 3 restrictions to what GOMS models can be used for:

  1. there must be a way to do the task in question

  2. the task must be able to be 'routinely done' by a skilled expert

  3. the GOMS analyst must start with a list of top-level tasks or user goals (provided from external sources outside of GOMS)

GOMS Models:

  1. KLM:  Keystroke Level Model (Card, et. al)

    • this is the simplest GOMS:  turns a simple action into a sequence of steps for analysis

    • Time of execution = T(execute) = Tk + Tp + Td + Tm + Tr
      Operators:  K (keystroke); P (pointing); H (homing); D (drawing); M (mental preparation); R (system response time)

    • The main advantage is this allows for a very quick estimate of execution time with very little theoretical / conceptual baggage.  It is the most practical GOMS technique.

  2. CMN-GOMS:  Card-Moran-Newell GOMS

    • named this to differentiate itself from other GOMS models as they began to appear

    • based on the serial-stage model (see diagram above), not the Model Human Processor (MHP)

    • slightly more specified than general GOMS

    • Quantitatively:  predicts the operator sequence and execution time

    • Qualitatively:  focuses attention on methods to accomplish goals.  Similar methods are easy to see.  Unusually short or long methods jump out and can spur design ideas.

    • While the KLM has no explicit goals, CMN-GOMS does.  The output of CMN-GOMS is in program form, so it is executable.

  3. CPM-GOMS:  Cognitive, Perceptual, Motor GOMS (Bonnie John)

    • based on the MHP (Model Human Processor) - Cognitive, Perceptual and Motor Operators can run in PARALLEL (subject to resource and information dependencies)

    • based on parallel multiprocessor stage model of human information processing

    • CPM also stands for Critical Path Method, since the critical path in a schedule (of parallel processes) determines the overall execution time.

    • Quantitatively:  predictions of performance times can be read off the chart of CPM-GOMS

    • Qualitatively:  analysis of what portions of a design lead to aspects of the performance are easy once the models are built

    • as with KLM, selection rules are not explicitly represented in the chart because the chart is just a trace of the predicted behavior

    • This is the only model representing parallelism, so several goals can be achieved at one time, which is characteristic of expert performance of a task

Human Information Processing


Chapter 5:  Notational Systems- The Cognitive Dimensions of Notations Framework


Think of the point of view of the system creators (practice side, not so theoretical). 

System designers try to incorporate new HCI Theories into their designs, but have learned that some are more practical than others.  They want to use theoretical knowledge, but prefer things like checklists which often aren't too theoretical.  The authors believe that it will not be possible to deal with these new notational systems by creating new checklists.

Especially with user testing- many of the methods that have been suggested such as user testing (laborious, expensive, in artificial labs) and Predictive models (expert-done calculations that predict time to complete a task on a finished system) are very expensive / time consuming to use

HCI has generated several approaches that offer suggestions for redesign, but they focus on representation and require extensive, detailed modeling.  There is no ONE approach that addresses all types of activity that can lead to constructive suggestions for improving the system or device, that avoids details and allows for the identification of similar problems down the road.  In HCI, no one model is perfect- each one has its own limitations, which is why we should know about the cognitive dimensions framework.

Alan Blackwell and Thomas Green propose a set of cognitive dimensions framework that will allow researchers and designers to 'talk together' (DISCOURSE) about evaluation.  It is much lighter and easier to use than the above two main types of evaluation.

Cognitive Dimensions Framework

- not a method; it is a set of discussion tools; used by the designers themselves


- main process is to DISCUSS TRADEOFFS of design decisions among designers by using a SHARED VOCABULARY

- Cognitive Dimensions Framework suggests a handful of basic 'Notational Dimensions' in the shared vocabulary:  VISCOSITY, HIDDEN DEPENDENCIES , ABSTRACTION LEVEL, PREMATURE COMMITMENT


Notational Dimensions (the SHARED VOCABULARY)
- each dimension describes an aspect of an information structure that is reasonably general

Evaluation using the CD Framework

There are two steps to evaluation using the cognitive dimensions framework:

  1. Decide what generic activities a system is desired to support.  Each generic activity has its own requirements in terms of cognitive dimensions.
  2. Scrutinize the system and determine how it lies on each dimension.  If the two profiles match, all is well.


What is nice about the CD framework is that it eliminates design maneuvers in which one dimension is 'traded-off' against another.  However, there are certain relationships:  such as a way to reduce viscosity is to introduce abstractions.  Abstractions might reduce viscosity and increase visibility...


Because CD Framework is so general, it has been used to structure questionnaires to get user-feedback on certain aspects (notational dimensions) of a system.  Because the notational dimensions are very general, feedback can be very useful and may be unanticipated by the designer (a good thing).


Chapter 6:  Users' Mental Models
- stems from cognitive psychology
- why?  because understanding users' mental models can enrich our understanding of the use of cognitive artifacts

Cognitive psychologists think the UMM is at the heart of understanding HCI- reasoning behind what we do

However, the term is over used in so many different ways that it has lost its usefulness

Still, mental models is a very important and useful construct, with many areas that can still be researched

Cognitive psychology foundations to mental models:

Idea 1:  Mental content vs. cognitive architecture:  mental models as theories

Idea 2:  Models vs. Methods:  mental models as problem spaces

Idea 3:  Models vs. Descriptions:  Mental models as Homomorphisms

Idea 4:  Models of Representations:  Mental models can be derived from language, perception or imagination

Idea 5:  Mental representations of representational artifacts

Idea 6:  Mental models as computationally equivalent to external representations

Case studies:

What are 'yoked' state spaces? 

'yoked state spaces' can motivate an informal analysis of the fit between the representational capacities of a device and the purposes of a user (calendar / appointment diary example)

support the process of internalization - computationally equivalent mental versions of external representations


Chapter 7:  Exploring and Finding Information

How do people forage for information?  People are informavores (George Miller):  we are organisms hungry for information about the world and themselves.  This chapter draws much from evolutionary theory.

Two main concepts of this chapter:

Information-Foraging Theory

Information Scent

Both ideas evolved in reaction to the technological developments associated with the growth of globally distributed, easily accessible information, and the theoretical developments associated with the growing influence of evolutionary theory in the behavioral and social sciences.

Following the metaphor:  What do we do when we hit a roadblock (metaphorically like hitting a stream, where we loose our 'information scent')?  We need to search around to pick up the scent again- much of the probability of success depends on the interface design (whether it's good or not- and can lead us in the right direction:  provide clues as to which choice to choose next, eliminate bad paths, etc.)  If we loose track of our 'scent' and accidentally pick up the wrong scent, we loose trust in the system.  Also, how far down the wrong path (scent) will we go before we realize we are on the wrong path?

Adaptation vs. Exaptation

this chapter's approach sees users as being adaptable.  Users are complex adaptive agents who shape their strategies and actions to be more efficient and functional with respect to their information ecology.

Adaptationist approaches became mainstream in the 1980s, as a reaction to ad-hoc models on human cognition (cognitive & perceptual tasks).  This is in contrast to mechanistic approaches of the time, such as the MHP (Model Human Processor, Card, et. al 1983)

Adaptationist approaches reverse-engineer the problem- examining what environmental problems are being solved and why cognitive and perceptual systems are well adapted to solving them

Exaptation looks at how we as humans are capable of adapting to solve similar problems (a way of generalizing our knowledge:  we can solve similar but different problems by observing themes in the problems)

Extending the Information Foraging Theory Metaphor:

Optimal-Foraging Theory

Goal is to optimize the information we receive to be the most rich, relevant to our information needs.  Optimization models include three major components:

In general, all tasks can be analyzed according to the value of the resource currency returned and costs incurred.  Classified as Resource costs (cost to get it) and Opportunity costs (potential benefit of it)

Scatter / Gather

As Designers:  we want to design systems that provide the richest information, with the least cost to access it (easy information retrieval) - can be very useful in web design and search engines

Current Status of IFT:  Internet Ecology is being studied- looks at complex global phenomena that yield predictions on Internet usage and distribution of users over web sites


Chapter 8:  Distributed Cognition

Almost every work situation requires someone working with other people

HCI hasn't had a way into this problem because cognitive theory tells us little about social behavior, and social science is hard to apply directly to design

Ways in which people use tools (artifacts) to support their goals is poorly understood

Distributed Cognition grew out of a need to understand how information processing and problem solving could be understood as being performed across units LARGER than the individual
- cognitive scientists (psychology) don't have to abandon their background to understand this:  we just shift from focusing on 'information in the head' to 'information in the world' - examine things 'in the wild'
how to understand how intelligence is manifested at the systems level rather than the individual level

Designing Collaborative Technologies

CSCW (Computer Supported Collaborative Work) shifts the focus of HCI from the user to the group (multiple, codependent users and their social network)

we focus on the problem solving of the group as a cognitive problem

we focus on distributed cognition within a context - drawing on actors and other features within the environment that allow problem solving (socially distributed cognition)

the goal of analysis:  describe how distributed units are coordinated by analyzing interactions among individuals, the representational media used, and the environment that the activity takes place

shift from traditional HCI (micro-structure level) to macro-structure level (ecological design - a turn to the social)

this has led ethnography - an anthropological approach to collecting data about the problem domain - to become a central feature of CSCW (see Ch. 13)


Distributed Cognition and Computing

Doing DCog:  Cognitive Ethnography

Unit of analysis:  the functional system (individuals, artifacts, and their relations)

Look for information-representation transitions that result in the coordination of activity and computations:


Chapter 9:  Cognitive Work Analysis (sometimes called sociotechnical work analysis)

Cognitive engineering - the analysis, modeling, design and evaluation of complex sociotechnical systems
- first coined by Norman after Three-Mile Island crisis, helped launch the field (how to design human-machine systems so they are safer / more reliable)
- goal was to design systems with good user interfaces, so that in unanticipated situations, the user would be able to safely handle the physical system

Cognitive Work Analysis (CWA) - way of analyzing human cognitive work - describes the forces that shape human cognitive work in complex real-time systems (where humans control complex physical processes).  The goal is to lead to systems that better support human-operator adaptation when operators are confronted with unanticipated variability.  CWA is a multidisciplinary approach to cognitive engineering.

Often, the human operator is very separated from the actual process he is managing, due to a poor interface.  CWA is an approach to cognitive engineering that aims to help find better ways of connecting the operator with what he is managing (physical system behind the technology) - so in unexpected situations, adaptive behavior (by the operator) will be successful.

Ecological Interface Design (EID) - subset of CWA - a set of principles to guide interface design (influenced by ecological psychology).  Said to be one of the most useful products of CWA.  Makes use of WDA and WCA (see below), as well as some ecological concepts.

formative model - an approach that describes the requirements that must be satisfied so that a system can behave in a new, desired way.  CWA is a formative approach to work analysis.

Phases of Analysis in CWA:  Based on Figure 9.3, p.230
 - each narrows down the action possibilities further from the previous phase

1.  Work Domain Analysis (WDA):  find purpose and structure of work domain; often represented in abstraction-decomposition diagrams and abstraction hierarchies
 - constraints (physical & purposive) within which activity takes place (NOT the activity itself, though)
 - often 5 levels of abstraction:  functional purpose, abstract function, generalized function, physical function, and physical form (see Figure 9.7 on p. 242 for an example)

2.  Control Task Analysis (CTA):  find what needs to be done in the work domain so that it can be effectively controlled; often represented in maps of control task coordination, decision ladder templates
 - continues to narrow down the 'dynamic space of functional action possibilities' by defining constraints that must be satisfied when work functions are coordinated over time and when effective control is exercised over the work domain

3.  Strategies Analysis (SA):  find ways that control tasks can be carried out; often represented in information flow maps
 - 'HOW control tasks can be done' is analyzed (we don't care by whom)
 - focus on general classes of strategies and their intrinsic demands

4.  Social-Organizational Analysis (SOA):  find who carries out work and how it is structured; often represented by annotating the information flow maps of #3
 - focus on the division and coordination of work (the content (information passed between actors)), and the social organization of the workplace (the form (behavioral protocol of communications))

5.  Worker competencies analysis (WCA):  find the kinds of mental processing supported; often represented by skill-based, rule-based, and knowledge-based behavior models

CWA was influenced by systems thinking and ecological psychology.  Both emphasize that the human-environment system needs to be the unit of analysis, with the environment being a primary unit of analysis in an actor's goal-oriented behavior.

Systems Theory:
- the whole is more than the sum of its parts (study the whole environment)
- study the relation between parts, not the properties of the parts
- study cybernetics - open vs. closed systems - and how outside disturbances affect the system

Ecological Psychology:
- CWA wants to build 'ecological information systems' that can be operated closer to the ease which the natural world is navigated
- EID (ecological interface design) is an approach to building interfaces using principles of CWA
- "ecological science rests on the principle that systems in the natural and social world have evolved to exploit environmental regularities" - Rosson & Carroll
- ecological psychology is the study of the way organisms perceive and respond to regularities in information
- key concept #1:  environments and information should be described in terms that reveal their functional significance for an actor rather than being described in objective actor-free terms
- key concept #2:  the affordances of an object are the possibilities for action with that object, from the point of view of a particular actor
- key concept #3:  the actor uses direct perception, which proposes that certain information meaningful to an actor is automatically picked up from the visual array

The CWA System Life Cycle (SLC):  requirements definition, modeling and simulation, tender evaluation, operator training, system upgrade, and system decommissioning.  This area may see much development in the next decade.

At the end of the chapter, many case studies are presented.  They fall into 2 categories:  display design, and evaluations of human-system integration.


Chapter 10:  Common Ground in CMC:  Clark's Theory of Language Use

HCI has come to encompass technologies that mediate human-human communication (chat, etc)

Production + Comprehension = Communication

Clark's theory of common ground:  we use existing common ground to develop further common ground and hence to communicate effectively

Grounding:  process of making sure that another person sufficiently understands you.  If not- use grounding.  Often facial expressions (non-verbal behavior) or questioning serves to notify us if other person needs more information (grounding)

Summary:  Clark's theory of language use is applicable to CMC (computer mediated communication).  The usage / application of this theory to designing systems will be more evident as time goes on- and will likely be very useful to new systems supporting video conferencing, asynchronous and synchronous communication, etc.  However, a very useful set of guidelines based on this has yet to be developed.


Chapter 11:  Activity Theory

History / Foundation for Activity Theory:

A basic perspective on HCI that came from Scandinavian roots in the early 1980's

an 'Action Research Approach' that focused on active cooperation between researchers and 'those being researched' - researchers entered into an active commitment to improve the situation of those they researched

roots in social psychology, industrial sociology and critical psychology.  Influence by the introduction of the personal computer, moving away from the mainframe.

There were many problems with existing HCI and system design at the time.  Much of the system design being done at the time consisted of heavy task analysis, with consideration of a generic novice user working alone.  A new theoretical foundation was needed!

Foundations of Activity Theory:

Activity theory shares with ecological psychology the attempt to move away from the separation between human cognition and human action, and an interest in actual material conditions of human acting.  Activity theory differs by adding a notion of motivation.  Activity theory values hierarchical analysis, task analysis, etc.  Activity theory takes place on all levels at the same time, not in sequence (see below).  Activity theory moves away from the generic user.

Unit of analysis in Activity Theory:  motivated activity.  This is mediated by socially produced artifacts (tools, language, representations, etc.)

Activity is mediated through a computer or other tool

Activity theory does not assume a boundary between internal representations and external representations, like cognitive science.  Activity theory has a basic feature of unifying consciousness and activity.

Development:  The most distinct feature of Activity Theory is development.  When compared to other materialist accounts in computer science, the focus is on development.

Nardi's Five Principles of Activity Theory:

Mediation is folded in with each of the other four principles, resulting in four categories of concerns:

Focus and Focus Shifts:

Activity theory starts with a perspective / point of view - which yields the objects to work with.  However, the focus shifts that indicate the dynamics of the situation are the main point of concern in the analysis.

(see Index page for more on Focus Shifts)


Nardi suggests that activity theory is a powerful descriptive tool rather than a predictive theory.  Carroll somewhat disagrees- because in this chapter concrete techniques were presented that show how it can be used to focus on computer-mediated activity (a.k.a. HCI)


Chapter 12:  Applying Social Psychology Theory to Group Work

              Study _______________________
                   /                      /|
Approach/                      / |
                 /                      /  |
         Build  /______________________/   |
                |                      |   |
     Technical  |                      |   |
Infrastructure  |                      |   |
               F|      Variations      |   |
  Architecture O|       in CSCW        |   |
               C|       research       |   |
   Application U|                      |   |
               S|                      |   |
          Task  |                      |   /
                |                      |  / 
        People  |                      | /  
  Group Size
Individual ... Group ... Society

CSCW (Computer Supported Collaborative Work) is a subfield of HCI that aims to build tools to help group work, learning and playing.  It examines how groups incorporate tools into their routines and the impact of CMC on group processes and outcomes.

Two main goals of CSCW:

Elements of an input-process-output model of groups:





Production Outcomes:  multidimensional

Inputs:  both inputs and processes that group members use will determine the success of the group

Interaction Process:  the way group members interact can directly influence group outcomes and mediate the impact of inputs on the group. 

Process Losses

Anonymity:  a way to offset social pressures, but can cause social loafing (trade-off)

CSCW researchers turn to the social-science literature outside their own field, and often consult this- such as ethnographic research- than experimental social psychology

Carroll says that CSCW research has been underexploited- a lot due to mismatching goals and values between HCI/CSCW research and social psychology research

social psychology has provided a rich body of research and theory about principles of human behavior, which should be applied to the design of HCI applications- especially those supporting multiple individuals who are communicating or performing a collaborative task


Chapter 13:  Studies of Work in HCI

Jonathan Grudin argued in 1990 that HCI was passing from the 4th stage to 5th stage:  from "a dialog with the user" to a "focus on the work setting"

This leads into the topics of this chapter, which get into ethnography, situated action, and ethnomethodology.


Many factors precipitated the adoption of CSCW:

Overview:  A Paradigmatic Case

Scientific Foundations


Situated Action


Conversation Analysis

Turn taking in organization is organized by participants on a moment-by-moment basis.  Turn taking is an interactional phenomenon, relating to multiple participants and organized as a collaborative matter.

Conversation analysis has developed as the study of turn taking in conversation.  It underpins some of the ways to study work, such as in the development of natural language interfaces.

Ethnomethodological Studies of Work

The second major preoccupation of ethnomethodology is its interest in work.  Ethnomethodological studies of work attempt to examine domains of work in order to understand what are the particular constitutive features of work, and how in their actions with one another, people are recognizably engaged in doing their work.



In the past 10 years, there has been a shift in HCI from the user to the social world (work setting) in systems design.  The door has been opened to study the work setting, and information being gathered from the work place will allow us to design better systems.


Chapter 14:  Upside-Down Algorithms - Computational Formalisms & Theory

This chapter is about how to gain insight into the context of HCI design and evaluation by using theoretical concepts (computational theory) and methods (formal methods).  We must understand our raw material (the computer itself) and that drawings, sketches, etc. are part of the design process.

Formalism is about being able to represent things in such a way that the representation can be analyzed and manipulated without regard to the meaning.  An example of this is State Transition Networks.

Key Features of Formal Descriptions:

Detailed Description:

Reasons for using Formal methods in HCI:

  1. to analyze the system to assess potential usability measures or problems
  2. to describe the system in sufficient detail so that the implemented system is what the designer intends
  3. the process of specification forces the designer to think about the system clearly and to consider issues that would be missed

Formal Modeling can be done for single user systems or for cooperative work

Case study:  flowcharts of the human-computer dialogue:  work well because they are simple formal methods.  Why:


There once was a time when every computer method had to be formal to be respectable.  However, formal models bred problems over time, and grew to the point where they were written off.  However, we should not write off formal methods - look at how successful UML diagramming has been.  Often using a formal model can be beneficial- especially those that projects simple in nature.

The web is an example of a rapidly growing distributed system.  Things like bookmarks and the 'back' button are stupid- it restarts the application in the middle of a process!

The problem and challenge of formal methods:  whenever we capture the complexity of the real world in formal structures, whether language, social structures, or computer systems, we are creating discrete tokens for continuous and fluid phenomena.  In doing so, we are bound to have difficulty (it is impossible to capture everything, in minute detail to a program, process, etc.).  However, it is only in doing these things that we can come to understand, to have valid discourse, and to design.  (think of Formal Models as 'naive physics' - as simple models that can work quickly / easily, but do not accurately represent reality)


Chapter 15:  Design Rationale as Theory

This chapter aims to show how reflective HCI design practices (involving design-rationale documentation and analysis) can be used to: 

  1. closely couple theoretical concepts and methods with the designed artifacts that instantiate them
  2. to more closely integrate theory application and theory development in design work
  3. to more broadly integrate the insights of different technical theories

Design rationale contributes to theory development in HCI in three ways:

  1. it provides a foundation for ecological science in HCI by describing the decisions and implicit causal relationships embodied in HCI artifacts
  2. it provides a foundation for action science in HCI by integrating activities directed at description and understanding with those directed at design and development
  3. it provides a framework for a synthetic science of HCI in which the insights and predictions of diverse technical theories can be integrated

Design Rationale

Applying theory in HCI design involves mapping concepts across domain boundaries, and directing descriptions and analysis to prescriptions for intervention.

TAF:  Task-Artifact Framework

Design Rationale:  Three scientific foundations:

  1. Ecological science:  rests on the principle that systems in the natural and social world have evolved to exploit environmental regularities. 
    • HCI can be developed as an ecological science at three levels:
      1.  taxonomic science
      2.  design science
      3.  evolutionary science
  2. Action science:  a principle to research that closely couples the development of knowledge and the application of that knowledge.  Integrates the traditional scientific objectives of analysis and explanation with the engineering objective of melioration.
  3. Synthetic science:  the design rationale that surfaced during a design project can be grounded in existing scientific theory, or can instantiate predictions that would extend existing theory

Detailed Description:

Case Study:  MOOsburg

The design rationale associated with an interactive system can be evaluated and refined.  In the case of MOOsburg, prototypes are still being developed, analyzed, used and refined.

When a design rationale is generalized, the hypothesized causal regularities contribute to theory building.  Design rationale supports ecological science at three levels:

When the generalized rationale is grounded in established scientific theory, it serves as an integrative frame within which to understand and further investigate competing or complementary concerns (synthetic science)

In general, scenarios and design rationale specify a shareable design space that can be used to raise, discuss, and arbitrate widely varying theoretical prediction and concerns.



Jump to Chapter:  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15