02 Multitasking

• Task switching

• Interruption (Memory for Goals)

• Multiple Tasks (Threaded Cognition)

• Prospective Memory

• Longer time frames of multiple tasks (STOM model)

• UV: 2 discrete tasks that alternate (letter task, digit task)

• AV = switching cost

• Limitationen: Labor, nur Kognition, fehlende "real world implication”

• Task switch = switch von letter zu digit task

• Distraction/attention request

• Interruption lag

• Interruption

• Resumption lag

visual or auditory event that observably captured the attention of the participant and delivered some information

time between alert and actual begin of secondary/interruption task

time that secondary/interrupting task is attended (retention interval)

time to resume primary task. Longer resumption lags indicate a more difficult resumption process, compared to shorter resumption lags

• Adaptive Control of Though-Rational (ACT-R) model to explain cognitive processes involved in deferral and resumption of task goals

• goals are stored in long-term memory; goal with highest activation will be retrieved and directs behavior; goal activation decays over time

• goal activation is elevated by:

  • strengthening (rehearsal)

  • priming (attention towards memory items or contexts that are associated with goal)

• longer interruptions lead to longer resumption lags

• log function of decay; rapid rise of lag for interrupting length of 3-13s and asymptotic at 45s

• more demanding interruption task resulted in longer resumption lags (due to less rehearsal during interruption)

• forced interruption lags decrease the resumption lag (more/better encoding, higher goal activation weil mehr Zeit)

• visibility of primary task during the interruption, audio interruption during a visual primary task and no context change during interruption decrease the resumption lag

• only one task

• what constitutes a “single task goal”?

• the primary task (that needs to be resumed) is always in central visual attention at the end of the interruption -> realistisch?

• einzige AV: Resumption lag (-> no forgetting to resume task, almost no errors at task resumption?)

• Fehler treten nicht auf

• vor allem Laborstudien

• Threaded cognition theory

  • Account for concurrent Multitasking performance for two or more tasks

• ACT-R cognitive architecture

  • provides theory and computational framework

• Memory for goals theory

  • account fo task interruption and resumption (sequential multitasking)

• task threads instead of goal activation

• multiple active tasks (threads) to execute simultaneously across the ACT-R processing modules

• threads are coordinated by procedural resource and executed by resource modalities

• resource seriality assumption

  • threads can operate in parallel if they do not require the same processing (each module can serve only one task at a time)

• greedy/polite polics

  • if a module is not in use, any of the threads can claim it (greed)

  • however, once a thread is done with a module, it releases it immediately (politeness)

• the declarative module can lead to interference when two tasks both use memory for standard memory retrievals or for retrievals of task instructions

• the problem representation module can lead to interference when two ore more tasks require complex representations that mus be swapped out for each task switch

• the procedural module (= the central cognitive bottleneck of the system) can lead to interference when two tasks attempt to start using the same module at the same time

• dual choice

• pursuit tracking

• driving and driving distraction

• interruption and task resumption

• uvm

• the primary task (which needs to be resumed) is always in central visual attention at the end of the interruption

• einzige AV: resumption times (no forgetting to resume task, almost no errors)

• model requires detailed input

• mainly lab studies

• memory processes involved in remembering future tasks

• resumption of interrupted task can be considered als prospective memory tasks

• Encoding-Pausen, Encoding-Reminder, Retrieval-Pausen und Retrieval-Reminder verbessern Wiederaufnahmeanteil (s. Grafik)

• computation multi-attribute decision model that predicts the decision aspect of sequential multitasking

• overload situations where concurrent performance of tasks is impossible -> need to decide which task should I do?

• what influences this decision?

1. Sensory properties of a task (Salience)

2. Effort of task switching (Effort): switch avoidance

3. Task priority (Value)

4. Task Interest

5. Difficulty of a task attribute

• Communication behavior

• Artefakte (Assistive or everyday technology Handys zb)

• planning and discretionary behavior

• affective component

• …

• detaillierte Models und viele empirische Studien verfügbar

• die meisten Studien sind Laborstudien, wenig real life

• viele mögliche Anwendungen / Anwendungspotenzial

• support for Memory for Goals:

  • decay (length of interruption)

  • priming (context change)

• does NOT support Memory for Goals:

  • factors that are based on rehearsal

  • nurses cann rehears, but did not actually rehears primary task

• dialing and driving: comparing different dialing methods and modeling the primary and secondary task performance

• UVs:

  • dialing input (manual vs voice)

  • number of actions for dialing (speed vs full)

• AVs:

  • time for dialing

  • driving performance

• declarative knowledge: individual units of factual knowledge (chunks)

• procedural knowledge: process and skills for achieving a goal (set of rules that fire when conditions are satisfied)

• perceptual modules

• motor modules

• control of vehicle (lateral steering and longitudinal speed)

• monitoring external environment

• decision making

• based on task analysis

• standard elements of ACT-R

• driving + dialing model (two subgoals)

• if dialing is required, driving model cedes control to dialing model with p=0.5 after each control action

• the dialing model returns control when a perceptual or motor action has been completed

• N = 11

• Nissan based driving simulator

• one lane

• constant speed (~100kmh)

• 8*4 calls, 20s between calls

• driving performance: during use + 5 seconds after

• the full-manual interface had large significant effects on driver performance, the speed-manual interface had small significant effects, and the voice interfaces had no significant effect

• the speed-manual interface required the least-time, followed by the speed-voice interface, the full-manual interface and finally the full-voice interface

• simulated and not real driving

• single lane, no traffic, constant speed

• keyboard simple and no feedback

• model limitation: individual differences and learning not adressed (zB “dial by feel”)

• Do different mobile information displays cause different effects on the resumption of the primary task?

• AV = resumption lag

• first study in laboratory setting

• UV1 Mobile Information Display

  • visibility/availability of primary task during interruption task reduces resumption lag

  • Tablet > Opaque HWD = HWD glance mode > HWD AR mode

• UV2 Interruption Length

  • longer interruptions result in longer resumption lags

  • 5s < 15s < 30s

• Display x Length Interaction

  • no effect of length for HWD AR mode

• Interaction interval significantly shorter than resumption lag -> interruption was distracting

• no main effect of UV1 (Mobile Information Display)

• Main effect of UV2 (Interruption Length): 5s < 15s = 30s)

• no interaction effect between UV1/UV2

• to investigate the use of a combination of psycho-physiological sensors to automatically identify the interruptibility of a knowledge worker in a real-world working context

• to investigate the correlation between interruptibility, mental load, and interruption lag (and resumption lag)

• Electroencephalograph (EEG) and eye blink data

• Electrodermal activity (EDA), skin temperature, blood volume pulse, interbeat interval, heart rate

• interruption on tablet

  • negotiated interruption (selbst entscheiden wann mans angeht)

  • one arithmetic task (1+2)

  • 5-point Likert scale questions about disturbance, interruptibility and mental workload at time of distraction

• Lab study

  • N = 10

  • 60 min programming in Eclipse IDE

  • random interruptions

• Field study (office of software company)

  • N = 9

  • 120 min of regular office work

  • random interruptions

• Distraction: Tablet played sound, changed from black to white, start button visible

• Resume Primary Task: Click on keyboard/mouse (lab: capture software, field: direct observation)

• Edit lag: time from first click to first edit

• classifications: disturbance, interruptibility, mental workload

• classification in interruptible vs. not interruptible:

  • Lab: 91.5%, Field: 78.6%

• classification in five interruptible states:

  • Lab: 43.9%, Field: 32.5%

• interruption lag lab 30s and field 45s (!)

• large positive correlations between perceived disturbance, interruptibility and workload (all r > .7)

• significant correlation between interruptibility and interruption lag

  • lab r=0.382, field r=0.282

• significant correlation between resumption lag and interruption lag in the field study, r=0.275 (schlecht für Theorie)

• psycho-physiological sensors of automatic identification of interruptibility worked well in lab and field

• Memory for Goals theory not well supported by results

• Limitation: artificial interruption