GOMS as a Simulation of Cognition Frank Ritter, Olivier Georgeon 30 oct 2017

Cognitive architectures "Brain emulator" Simulates the computation we think the brain does (for doing a given task…). Provides structures to store symbols Provides instructions to manipulate symbols Hypothesis: Goal-driven, Problem-solving, symbolic computation. A way to combine what we know about users and apply.

GOMS Architecture Explain each part. Note that there is some parallelism, but that much of most tasks are simply serial. You can also point out that the previous chapters provide details on most of these boxes.

Example: Check email Procedure check_emails Goal Login Goal read_email Goal memorize_sender's_name Goal read_email_body If Needed(), Goal replay_email Goal reply_email(x) Selection rule Select_appropriate_formulation(x) Goal reply_email_friend Operator Type (Hi <x>) … Goal reply_email_familly Operator Type (Dear <x>) Loop to read_email / Repeat Goal read_email Goal: Logout Note: Needed ()’s will be translated as a mental op, but is usually more complex Refer them to Kieras’ manual, which is a good manual.

Mental operations and contents Visual buffer Long term memory Perceived_item: John (John, relation, friend) (Jack, relation, family) (email, name, John) Not a complete list, but a sample of what is used by a use. Goals Selection rule Reply_email_friend Reply_email_family Memorize_sender's_name … If Perceived_item = X And (X, relation, friend) Then Process goal reply_email_friend

How To Use GOMS Analyze hierarchical structure of a task coarse analysis focuses more on the cognitive structure of a task fine analysis focuses more on the structure imposed by the specific interface design Analyze alternative methods of interaction Assign operators to base level goals Assign times to operators Sum the operator times to get sum for task Look for excessive time Look for duplicated time Look for wasted time Look for non-parallel construction on parallel tasks A-D go on a ways as the analysis have lots of ways to show how to improve Ahh, look for where the system response time is killing you

GOMS Operator Times 280 ms 1,500 ms 300 ms 230 ms 1,200 ms 25,000 ms Press key on keyboard 280 ms Use mouse to point to object on screen 1,500 ms Move hand to pointing device 300 ms Move eyes to location on screen 230 ms Retrieve item from memory 1,200 ms Learn a single step in a procedure 25,000 ms Select among methods More available in FDUCS, GOMSL and CM&N

Summary of GOMS A method to describe tasks and how a user performs those tasks with a specific design bridges task analysis with a specific interface design assumes error-free, goal-directed, and rational behavior Views humans as information processors small number of cognitive, perceptual, and motor operators characterize user behavior To apply GOMS analyze task to identify user goals (hierarchical) identify operators to achieve goals sum operator times to predict performance

KLM Times To do the KLM, keep track of 80-1200 ms 1,100 ms 400 ms Press key on keyboard 80-1200 ms Use mouse to point to object on screen 1,100 ms Move hand to pointing device 400 ms Mental OP: Retrieve item from memory 1,350 ms Select among methods 1,200 ms Slightly different numbers, because slightly different theory

M Heuristics placement Do all physical ops 0. Insert Mop before all K’s not part of strings (e.g., shift to new field) and P’s that select a command (not argument) Remove all Mops fully anticipated as part of commands (PMK-> PK) Remove all after first letter of a command (CR M K M K M K M-> CR M K K K K) Remove redundant command terminators (CR M command M CR -> CR M command CR) Don’t remove variable terminators (no change)

Conclusion: KLM and GOMS, an inexpensive way to improve about interfaces Cheap way to evaluate expert, error-free performance (GOMS) in interfaces Cheaper way to evaluate interfaces (KLM) If done fairly, typically tells you the best interface Typically tells where improvements are possible Less movements, less keys, less Mop Keep parallel and consistent Shift from mouse to keyboard for experts Apply Fitts law where appropriate to improve targets Change users (faster Ks) or hardware (better mouse or autokeypress, etc.)