cognitive models physical and device architectural Cognitive - PDF document

Cognitive models • goal and task hierarchies chapter 12 • linguistic cognitive models • physical and device • architectural Cognitive models Goal and task hierarchies • They model aspects of user: • Mental processing as divide-and-conquer – understanding • Example: sales report – knowledge produce report gather data – intentions . find book names – processing . . do keywords search of names database . . . … further sub-goals • Common categorisation: . . sift through names and abstracts by hand . . . … further sub-goals – Competence vs. Performance . search sales database - further sub-goals – Computational flavour layout tables and histograms - further sub-goals – No clear divide write description - further sub-goals goals vs. tasks Issues for goal hierarchies • goals – intentions • Granularity what you would like to be true – Where do we start? • tasks – actions – Where do we stop? how to achieve it • Routine learned behaviour, not problem solving • GOMS – goals are internal – The unit task • HTA – actions external • Conflict – tasks are abstractions – More than one way to achieve a goal • Error 1

Techniques GOMS • Goals, Operators, Methods and Goals Selection (GOMS) – what the user wants to achieve Operators • Cognitive Complexity Theory (CCT) – basic actions user perform s Methods • Hierarchical Task Analysis (HTA) - – decom position of a goal into subgoals/ operators Chapter 15 Selection – m eans of choosing between com peting m ethods GOMS example Cognitive Complexity Theory • Two parallel descriptions: GOAL: CLOSE-WINDOW . [select GOAL: USE-MENU-METHOD – User production rules . MOVE-MOUSE-TO-FILE-MENU . PULL-DOWN-FILE-MENU – Device generalised transition networks . CLICK-OVER-CLOSE-OPTION GOAL: USE-CTRL-W-METHOD . PRESS-CONTROL-W-KEYS] • Production rules are of the form: – if condition then action For a particular user: Rule 1: Select USE-MENU-METHOD unless another • Transition networks covered under rule applies Rule 2: If the application is GAME, dialogue models select CTRL-W-METHOD Four rules to model inserting Example: editing with vi a space • Production rules are in long-term m em ory Active rules: New working memory SELECT-INSERT-SPACE • Model working memory as attribute-value INSERT-SPACE-MOVE-FIRST (GOAL insert space) mapping: INSERT-SPACE-DOIT (NOTE executing insert space) INSERT-SPACE-DONE (LINE 5) (COLUMN 23) (GOAL perform unit task) (TEXT task is insert space) SELECT-INSERT-SPACE matches current working memory (TEXT task is at 5 23) (CURSOR 8 7) (SELECT-INSERT-SPACE • Rules are pattern-matched to working IF (AND (TEST-GOAL perform unit task) memory, (TEST-TEXT task is insert space) (NOT (TEST-GOAL insert space)) e.g., LOOK-TEXT task is at % LINE % COLUMN (NOT (TEST-NOTE executing insert space))) is true, with LINE = 5 COLUMN = 23. THEN ( (ADD-GOAL insert space) (ADD-NOTE executing insert space) (LOOK-TEXT task is at %LINE %COLUMN))) 2

Notes on CCT Problems with goal hierarchies • Parallel model • Proceduralisation of actions • a post hoc technique • Novice versus expert style rules • Error behaviour can be represented • expert versus novice • Measures – depth of goal structure • How cognitive are they? – num ber of rules – com parison with device description Linguistic notations Backus-Naur Form (BNF) • Very com m on notation from com puter science • Understanding the user's behaviour and cognitive difficulty based on analysis of • A purely syntactic view of the dialogue language between user and system. • Term inals • Similar in emphasis to dialogue models – lowest level of user behaviour – e.g. CLI CK-MOUSE, MOVE-MOUSE • Nonterm inals • Backus–Naur Form (BNF) – ordering of term inals – higher level of abstraction • Task–Action Grammar (TAG) – e.g. select-m enu, position-m ouse Example of BNF Measurements with BNF • Basic syntax: • Number of rules (not so good) – nonterm inal : : = expression • An expression • Number of + and | operators – contains term inals and nonterm inals – com bined in sequence (+ ) or as alternatives (| ) • Complications draw line : : = select line + choose points + last point – sam e syntax for different sem antics select line : : = pos mouse + CLICK MOUSE choose points : : = choose one | choose one + choose points – no reflection of user's perception choose one : : = pos mouse + CLICK MOUSE – m inim al consistency checking last point : : = pos mouse + DBL CLICK MOUSE pos mouse : : = NULL | MOVE MOUSE+ pos mouse 3

Task Action Grammar (TAG) Consistency in TAG • I n BNF, three UNI X com m ands would be described as: • Making consistency more explicit copy : : = cp + filename + filename | cp + filenames + directory move : : = mv + filename + filename | m v + filenames + directory • Encoding user's world knowledge link : : = ln + filename + filename | ln + filenames + directory • No BNF m easure could distinguish between this and a • Parameterised grammar rules less consistent gram m ar in which link : : = ln + filename + filename | ln + directory + filenames • Nonterminals are modified to include additional semantic features Consistency in TAG (cont'd) Other uses of TAG • consistency of argum ent order m ade explicit • User’s existing knowledge using a param eter, or sem antic feature for file operations • Congruence between features and • Feature Possible values Op = copy; move; link commands • Rules file-op[ Op] : : = command[ Op] + filename + filename • These are modelled as derived rules | command[ Op] + filenames + directory command[ Op = copy] : : = cp command[ Op = move] : : = m v command[ Op = link] : : = ln Physical and device models Keystroke Level Model (KLM) • lowest level of (original) GOMS • The Keystroke Level Model (KLM) • Buxton's 3-state model • six execution phase operators – Physical m otor: K - keystroking P - pointing H - hom ing • Based on empirical knowledge of D - drawing human motor system – Mental M - m ental preparation • User's task: acquisition then execution. – System R - response – these only address execution • times are empirically determined. • Complementary with goal hierarchies Texecute = TK + TP + TH + TD + TM + TR 4

KLM example Architectural models GOAL: ICONISE-WINDOW • All of these cognitive models make [select assumptions about the architecture of GOAL: USE-CLOSE-METHOD . MOVE-MOUSE-TO- FILE-MENU the human mind. . PULL-DOWN-FILE-MENU . CLICK-OVER-CLOSE-OPTION • Long-term/ Short-term memory GOAL: USE-CTRL-W-METHOD PRESS-CONTROL-W-KEY] • Problem spaces • Interacting Cognitive Subsystems USE-CTRL-W -METHOD USE-CLOSE-METHOD • compare alternatives: H[ to kbd] 0.40 P[ to menu] 1.1 • USE-CTRL-W-METHOD vs. • Connectionist M 1.35 B[ LEFT down] 0.1 • USE-CLOSE-METHOD K[ ctrlW key] 0.28 M 1.35 • ACT • assume hand starts on mouse P[ to option] 1.1 B[ LEFT up] 0.1 Total 2.03 s Total 3 .75 s Display-based interaction • Most cognitive models do not deal with user observation and perception • Some techniques have been extended to handle system output (e.g., BNF with sensing term inals, Display-TAG) but problems persist • Exploratory interaction versus planning 5