1. Gary MarsdenSlide 1University of Cape Town Case Study - Nokia 5110 We will try to put together what we have learnt to date by looking at a cell- phone, namely: Nokia 5110

2. Gary MarsdenSlide 2University of Cape Town Brief description Nokia 5110 is currently best selling GSM handset. Easy to use? –“Use your phone as you want. Send short messages, save names and numbers, select a new ringing tone - all with the press of a single key, the Nokia Navi™ Key.”

3. Gary MarsdenSlide 3University of Cape Town Design criteria I Affordance - not particularly relevant Mapping - arrow keys for menu Constraints - SIM card Visualising - one line menus, keypad lock, icons Memory - one line menus Knowledge - no on line help, unlike banana phone Conceptual model - dial number and press “Call” Turing test - come back to this Role Integrity - possible network / handset conflict

4. Gary MarsdenSlide 4University of Cape Town Design Criteria II Simplicity - Navi-key! Least astonishment - Ok so far. Modes - enter number in address book (not banana) Equal opportunity - some, but not optimal (ABC key, entry modes) Least effort - 6 key presses to change message centre number, 11 to change ring volume Feedback - Ok, generic messages problem Cognitive dimensions - not applicable

5. Gary MarsdenSlide 5University of Cape Town Humans as 2nd class citizens Thimbleby’s principle of equating humans to machines turns out to be very important Is it possible to write an algorithm to interact with the 5110? How efficient is the menu structure of the 5110?

6. Gary MarsdenSlide 6University of Cape Town Some definitions We wish to find how efficient interface access is Assume a perfect, naïve user (do not consider error recovery) Interested in access to functions, not their activation Count access in terms of key presses –5110 only interacts via keys, therefore there is a common ‘cost’ of interaction

7. Gary MarsdenSlide 7University of Cape Town Calculating cost A simple measure of usability can be gained as follows  button presses Number of functions The use of this type of simple model has been shown (Card, Carroll, Moran, Newell) Increase usability by restructuring access to functions, not increasing number of functions

8. Gary MarsdenSlide 8University of Cape Town 5110 in-depth 5110 has 74 functions of interest to us Arranged in a menu structure (see handout) –Average time: 8.2 presses –Max search: 110 presses –Max time: 14 presses

9. Gary MarsdenSlide 9University of Cape Town 5110 data structure Is it sensible to structure the interface around an arbitrary menu? As computer scientists, we know about data structures so can re-design to use a more efficient structure

10. Gary MarsdenSlide 10University of Cape Town Binary trees This time, we have 74 functions in binary tree Use left, right and select buttons –Average time: 5.4 presses –Max search: 152 presses –Max time: 7 presses Alphabetic –no worse than existing –functional clusters e.g. DIVERT

11. Gary MarsdenSlide 11University of Cape Town Who wants trees The previous two solutions assume a tree structure, but is this a good idea? Christopher Alexander’s classic paper - The City is not a Tree- argues against strict classification –e.g. Should volume be in “Tones” or “Phone Settings” Try some more data structures...

12. Gary MarsdenSlide 12University of Cape Town Linear lists Using a list, we remove navigation cognitive load - checking existence is easy Using alphabetic ordering –Average time: 37 presses –Max search: 74 presses –Max time: 74 presses Improvements through –Software: frequency ordering –Hardware: jog wheel

13. Gary MarsdenSlide 13University of Cape Town Hashing Using a technique known as hashing, we can exploit the letters on each key 7 PQR 8 STU

14. Gary MarsdenSlide 14University of Cape Town Benefits  The average number of key presses required to access an individual item was reduced from 8.2 using menus to 3.1 using the mapping algorithms.  Commands could be added dynamically (by the user or service provider) without the need to restructure menus.  The user was freed from modes – e.g. if they had started to enter a function name in the address book, the contents of the buffer could be remapped to the correct domain with one key press.  The solution promoted awareness of functions, as they could be seen scrolling down the screen.  It is possible to improve user recall of command names by providing a “thesaurus” of alternatives for each function name.  The system has a more manageable search space, permitting the user to back track and easily correct keying mistakes.

15. Gary MarsdenSlide 15University of Cape Town Limitations of the approach Assumes error free –longer the sequences, more likely the errors No assumptions about cultural factors No consideration of physical interaction form Time ignored –Time out etc. only compound problem Trivial? –Yes, but why is it not being done!

16. Gary MarsdenSlide 16University of Cape Town Conclusions Is any analysis done? Where / how does usability influence design? Do users care? Computing Science has a role to play –Many more algorithms available