Mo dels of the System Ov erview Standard F ormalisms soft - PDF document

Mo dels of the System Ov erview Standard F ormalisms soft w are engineering notations used to sp ecify the required b eha viour of sp eci�c in teractiv e systems In teraction Mo dels sp ecial purp ose mathematical mo dels of in teractiv e systems, used to describ e usabilit y prop erties at a generic lev el Status/Ev en t Analysis an example of an engineering lev el metho d dra wing on b oth formal mo delling and na � �v e psyc hology Human{Com puter In teraction, Pren tice Hall Mo dels of the System (1) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Relationship with dialogue Dialogue mo delling is link ed to seman tics. System seman tics a�ects the dialogue structure. But the bias is di�eren t. Rather than dictate what actions are legal, these formalisms tell what eac h action do es to the system. Human{Com puter In teraction, Pren tice Hall Mo dels of the System (2) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Iron y Computers are inheren tly mathematical mac hines. Humans are not. F ormal tec hniques are w ell accepted for cognitiv e mo dels of the user and the dialogue (what the user do ). should F ormal tec hniques are not y et w ell accepted for dictating what the system should do user ! for the Human{Com puter In teraction, Pren tice Hall Mo dels of the System (3) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

General computational formalisms Standard soft w are engineering formalisms can b e used to sp ecify an in teractiv e system. Referred to as formal metho ds Mo del based describ e system states and op erations � Z, VDM Algebraic describ e e�ects of sequences of actions � OBJ, Larc h, A CT-ONE Extended logics describ e when things happ en and who is resp onsible � temp oral and deon tic logics Human{Com puter In teraction, Pren tice Hall Mo dels of the System (4) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

The uses of SE formal notations F or comm unication � common language � remo v e am biguit y (p ossibly) � succinct and precise F or analysis � in ternal consistency � external consistency { with ev en tual program { with resp ect to requiremen ts (safet y , securit y , HCI) � sp eci�c v ersus generic Human{Com puter In teraction, Pren tice Hall Mo dels of the System (5) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Mo del based formalisms Mathematical coun terparts to common programming constructs Mathematics Programming t yp es sets basic t yp es basic sets constructed t yp es constructed sets records unordered tuples lists sequences functions functions pro cedures relations Human{Com puter In teraction, Pren tice Hall Mo dels of the System (6) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

The mo del based metho d Example: a graphic dra wing pac k age Line Rectangle Ellipse Unselect P oin ts are ordered pairs. N N == � P oint Shap es can b e of v arying t yp es. == ine j el ipse j r S hape ty pe l l ectang l e Human{Com puter In teraction, Pren tice Hall Mo dels of the System (7) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

7 More t yp e de�nitions A graphic ob ject is de�ned b y its shap e t yp e, width, heigh t, and cen tre p osition. S hape : ty pe S hape ty pe N : w idth; heig ht : centr e P t A collect io n of graphic ob jects can b e iden ti�ed b y a `lo okup dictionary' [ I d ] == ! S hape dict I d S hape Human{Com puter In teraction, Pren tice Hall Mo dels of the System (8) � c A. Dix, J. Finla y , G. Ab o wd and R. Beale 1993 Chapter 9

De�ning the state The system state con tains a dictionary of created ob jects and a set of selected ob jects. S tate : shapes S hape dict P : sel ection I d � dom sel ection shapes Initiall y , there are no shap es in the dictionary . I nit S tate 0 S tate 0 = fg shapes Human{Com puter In teraction, Pren tice Hall Mo dels of the System (9) � c A. Dix, J. Finla y , G. Ab o wd and R. Beale 1993 Chapter 9

De�ning op erations State c hange is represen ted as t w o copies of the state b efore | S tate after | 0 S tate � S tate S tate 0 S tate The op eration deselects an y selected U nsel ect ob jects. U nsel ect � S tate 0 = fg sel ection 0 = shapes shapes Human{Com puter In teraction, Pren tice Hall Mo dels of the System (10) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

In terface issues F raming problem \ev erything else sta ys the same" can b e complicated with state in v arian ts In ternal consistency do op erations de�ne an y legal transition? External consistency m ust b e form ulated as theorems to pro v e Clear for re�nemen t, not so for requiremen ts Separation of system functionalit y and presen tation is not explici t Human{Com puter In teraction, Pren tice Hall Mo dels of the System (11) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Algebraic notations Mo del based notations emphasise constructing an explici t represen tations of the system state. Algebraic notations pro vide only implicit information ab out the system state. Mo del based op erations are de�ned in terms of their e�ect on system comp onen ts. Algebraic op erations are de�ned in terms of their relationship with the other op erations. Human{Com puter In teraction, Pren tice Hall Mo dels of the System (12) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Return to graphics example t yp es State, Pt op erations : init S tate ! : mak e el l ipse P t S tate S tate � ! : mov e P t S tate S tate � ! : unsel ect S tate S tate ! : del ete S tate S tate ! axioms tate ; for all st S p P t 2 2 � 1. ete ( mak ipse ( st )) = ect ( st ) del e el l unsel 2. ect ( unsel ect ( st )) = ect ( st ) unsel unsel 3. e ( p; ect ( st )) = ect ( st ) mov unsel unsel Human{Com puter In teraction, Pren tice Hall Mo dels of the System (13) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Issues for algebraic notations Ease of use a di�eren t w a y of thinking than traditional programming In ternal consistency are there an y axioms whic h con tradict others? External consistency with resp ect to executable system less clear External consistency with resp ect to requiremen ts is made explicit and automation p ossible Completeness is ev ery op eration completely de�ned? Human{Com puter In teraction, Pren tice Hall Mo dels of the System (14) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Extended logics Mo del based and algebraic notations mak e extended use of prop ositional and predicate logic. Prop ositions expressions made up of atomic terms comp osed with p; q ; r ; : : : ( ; ) ; ^ _ ; : ) , etc. ; ; Predicates prop ositions with v ariables, e.g., p ( x ) and quan ti�ed expressions 8 9 . ; These are not con v enien t for expressing time, resp onsibili t y and freedom, notions sometimes needed for HCI requiremen ts. Human{Com puter In teraction, Pren tice Hall Mo dels of the System (15) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

T emp oral logics Time considered as succession of ev en ts Basic op erators: alw a ys 2 (G funnier than A) 2 ev en tually 3 (G understands A) 3 nev er 2 : 2 : (rains in So. Cal.) Other b ounded op erators: w eak er than p until q 2 stronger than p bef or e q 3 Human{Com puter In teraction, Pren tice Hall Mo dels of the System (16) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9

Explicit time These temp oral logics do not explici tl y men tion time, so some requiremen ts cannot b e expressed. Activ e researc h area, but not so m uc h with HCI Gradual degradation more imp ortan t than time-critical i t y Myth of the in�nitely fast mac hine Human{Com puter In teraction, Pren tice Hall Mo dels of the System (17) A. Dix, J. Finla y , G. Ab o wd and R. Beale � c 1993 Chapter 9