SEEV?
Salience
Effort
Expectancy
Value
Salience?
attention-attracting properties of a display
aufmerksamkeitserregende Eigenschaften eines Displays
Effort?
effort required to move attention between displays
benötigter Aufwand um die Aufmerksamkeit zwischen mehreren Anzeigen zu bewegen/switchen
Expectancy?
expectancy that information (change) will appear within a display
Erwartungshaltung, ob Information(-sveränderung) auf einem Display erscheinen wird
Value?
events appearing within a display may have value in their own right
relative importance of the task(s) served by that display
Events auf einer Anzeige können intrinsischen Wert (von sich aus) haben
auch relative Wichtigkeit von Aufgaben, die von einem spezifischen Display bereitgestellt werden
SEEV model / SEEV equation?
Attentional attractiveness of a displayP =
sS - efEF + exEX + vV
Kleinbuchstaben sind Gewichte
s and ef can be set to 0 -> creating a model that should guide attention in absence of potentially distracting salience and effort
Static vs Dynamic SEEV Model?
static: overall percentage dwell time on each display, but does not track the process of executing the scan path
dynamic: captures the sequence of fixations and not just their overall frequency (first-order Markov model)
salience: of the window scene is set larger than that of either the speed or heading displays (cause larger, color)
effort: is rated lower for the window and the speed than for heading (distance from center NOT CENTROID)
expectancy: assumptions that a) the window, depicting both heading and speed changes, will have a bandwidth that is the sum of the speed and heading displays and b) heading higher than speed (cause heading can be changed easier)
Value Berechnung?
value parameters involve the mediating variable of task importance and display for a task relevance
relevance = Frequenz
value = importance * relevance
Ergebnisvergleich hypothetische SEEV vs EV predictions?
-> macht Sinn, wenn man davon ausgeht dass für window display effort niedrig und salience hoch ist
Ergebnisse von SEEV/EV in verschiedenen Anwendungsfeldern?
-> alles recht hoch
Situation Awareness Model?
Situation Awareness?
Level 1: Perception of data and the elements of the environment
Level 2: Comprehension of the meaning and significance of the situation
Level 3: Projection of future states and events
Situation Awareness leads to..
a specific decision
The awareness-induced decision leads to..
the performance of the chosen action
The performance of the action which follows a decision leads to..
changes of the state of the environment/system, which in turn starts the situation awareness cycle again with the perception of the changed data/elements in the environment
Task and Environmental Factors?
influence situation awareness, decision-making and the performance of action
workload
stressors
system design
complexity
Individual Factors?
goals
preconceptions / assumptions
knowledge
experience
training
abilities
SEEV Limitations?
only visual attention
mainly focal attention (absichtlich gesteuerte Aufmerksamkeit)
effort and salience difficult to investigate (coding eye tracking data, modeling -> tedious)
parameters need to be assigned for specific/situations (kein generalisierbares Modellieren möglich)
Anästhesie Beispiel?
real cases weniger gute Predictions als in simulated cases
junior/senior nimmt sich nicht viel
“Messy real cases in the OR compared to “tidy” simulated case in a more sheltered environment”
Effort Bewegungen Beispielkategorisierung?
Cognitive Aids?
support user during the actual execution of a task
zB Checklisten, AR, Apps…
Hilfsapp bei CPR?
Probleme bei Medical Simulations?
simulation is limited
enough monitoring of patient vital signs in real settings?
SEEV Fazit Anästhesie?
Unterschiede zwischen Simulation und realer Umgebung in Bezug auf die visual attention distribution
including effort and salience did not improve model fit -> EV ausreichend
EV seems like a valid measure for situation awareness level 1 (perception)
EV model not comprehensive enough to predict higher situation awareness levels or decision/performance
EV model möglich als Tool um Technologie zu evaluieren
Beispiel aufwendige Modellierung für jeden Einzelfall?
Car drivers monitoring behavior Randdaten?
EV parameters estimaed by 20 experienced drivers vs 9 HF experts
20 VP (experienced drivers)
257 Manöver in SILAB (simulation software)
16 AOI classes
Car drivers monitoring behavior AOI classes?
von Experten bestimmt
Car drivers monitoring behavior expectancy/value parameters?
Car drivers monitoring behavior - Resultat 1: predicting own monitoring behavior?
Individual, experienced car drivers are bad in identifying the regions that they are looking at during an overtaking maneuver and also in modelling how they divide attention between these regions
bad in predicting their own monitoring behavior
r = 0.318 (wenig Übereinstimmung)
>75% der Blicke waren nicht Teil der AOIs
Car drivers monitoring behavior - Resultat 2: aggregierte Models?
aggregating the models of multiple experienced car drivers improves attention prediction results compared to an average individual experienced car driver
r=0.719
19% der Blicke waren nicht Teil der AOIs
Car drivers monitoring behavior - Resultat 3: aggregierte Fahrer vs HF experts?
HF Experten r=.394
-> aggregierte Fahrer sagen besser voraus
Car drivers monitoring behavior - Resultat 4: HF Experten aggregieren?
Car drivers monitoring behavior - Fazit?
16 AOIs
zeitlich effiziente Parameterschätzung (1h15Min)
Automatisiertes Data Processing!
average model scores of multiple domain experts for best predictions
Attentional Behavior in Public Display - Randinfos?
Fragestellung: is it enough to model human attentional behavior in public display?
Display auf Sportmesse
salience = positiv, konstant (extreme sports sind spannend)
expectancy = positiv, konstant (public displays sind bekannt für viel Infos)
value = positiv, konstant (Sport, freiwillige Anwesenheit -> Motivation intrinsisch)
effort??
Attentional Behavior in Public Display - Effort Annahme?
Attentional Behavior in Public Display - Behavioral Classes?
handlabeled / kodiert
0: No perception (display out of fov)
1: Selective perception (display in peripheral visual range)
2: Switched Attention (display actively in wandering gaze area)
3: Focused attention (conscious perception, distinct behavior change)
4: Sustained attention (lasting perception, fundamental behavior change)
Attentional Behavior in Public Display - Korrelation zwischen Parametern und Behavioral-Klassen?
Attentional Behavior in Public Display - Evaluation?
Averaging over single VP:
average 148frames/VP
bias for shorter samples (less frames means worse model fit)
MW=n.a., Md r=0.263
Overall correlation:
Annahme: 28.000 Frames = EIN User
Dann Korrelation zwischen 5 Klassen und Effort Model: r=0.749
Attentional Behavior in Public Display - Limitations?
kein safety-critical context
kein Kontrolltask zur Überwachung
Effort hat einen Unterschied gemacht
Ideal Scanning?
nur Expectancy und Value (EV)
Actual Scanning?
Alle vier Komponenten
Zuletzt geändertvor 2 Jahren