Human Factors Exam 3

Term
1 / 72
encoding
Click the card to flip 👆
Terms in this set (72)
encoding
measure the evidence for a signal
decision making
decide whether the evidence is strong enough to merit a yes judgment
signal detection theory is a way to separate _________ and ___________
sensitivity and bias
possible results for a trial:
hit, false alarm, miss, correct rejection
hit
if its a signal trial and you say a signal is present
ex: if there is a knife, and the TSA luggage screener stops it
false alarm
if its a noise trial and you say a signal is present
-ex: if there is no knife, and the TSA luggage screener stops it for inspection
miss
if its a signal trial and you don't detect it
-ex: if there is a knife, and the TSA luggage screener lets the bag pass without stopping it
correct rejection
if its a noise trial and you say there is no signal
-ex: if there isn't a knife, and the TSA luggage screener lets the bag go
what two factors determine bias?
signal rate and payoffs
as signal rate decreases...
b optimal (bias) becomes more conservative
as signal rate increases...b optimal (bias) becomes more liberalwhat are the two states of signal?threat present and response receivedwhat are the two states of noise?no threat and response not receivedB optimal (bias) =P (Noise) / P (Signal)increasing value (CR) or cost (FA) makes B optimal...more conservativeincreasing value (H) or cost (M) makes B optimal...more liberalas the likelihood of signal decreases...you should be more conservativeas the cost of missing a target and the value of recognizing a target increases...you should be more liberalpayoffs determine...optimal biasvigilancethe ability of organisms to maintain their focus on attention and to remain alert over prolonged periods of timevigilance tasks are signal detection tasks that...occur over a long period of timesignals are (3 things):intermittent, unpredictable, infrequentexamples of vigilance tasks:-radar monitoring -TSA baggage claim screening -quality control inspection -supervising complex systemvigilance tasks also known as...sustained attention taskNorman Mackworth's Clock Testclock hand jumps 1/sec, task is to detect double jumps; hit rate = ~85% at start of 2 hour watch, but fell 10-15% within half hour continued to decline after thatvigilance decrementdecline in detection rate that occurs over time -larger drop for difficult taskssimultaneous decrementoperator compares side to side stimulisuccessive paradigmoperator has to compare events to memory of a standardsuccessive tasks:-most likely to suffer form a vigilance decrement -far more mentally demanding, impose working memory load -cause more interference with a concurrent task -are more degraded by manipulations that make tase difficultexpectancy theorylow target frequency causes bias to become more conservative as watch continuesthe vicious circle of target expectancy:low P(target) leads to conservative B (bias)... conservative B (bias) causes subject to miss target... missed signals make subject underestimate P (target) -B (bias) is determined by subjective P (target) which might be inaccuratemotor schema-not all movement depends on visual feedback -highly learned skills when environmental uncertainty is low -said to be performed in an "open loop" fashion -feedback is proprioceptiveexamples of motor schematying you shoe, signing your name, typing -highly over learned skillslow attention demandtasks that have been practiced so much they are automated -require few resources to perform, so not disrupted much by a secondary tasksingle response selectionmay involve a complex sequence of discrete responses, but really only one response selection occurs to initiate or "load" the sequenceconsistency of outcomeproduct of motor schema consistent across executionscharacteristics of motor schemas-low attention demand -single response selection -consistency of outcome -not tied to repeated muscle movementswhat does "not tied to repeated muscle movements" mean relating to motor schemas?-representation store: relative timing of direction changes, relative position of direction changes -exs: singing name on different surfaces, different sizeswhat should e(t) be at whatever specified time?0the relationship between f(t) and u(t) is known as:control dynamics (force applied and movement)the relationship between u(t) and o(t) is known as:system dynamics (control position and system response)1st stage of tracking loophuman operator may notice a deviation (error, or e(t)) between where they are and where they want to be (ex: lane center)2nd stage of tracking loophuman operator applied force (f(t)) to an input device (e.g. steering wheel) to counteract this error3rd stage of tracking loopforce causes the control device to move (u(t)) -ex: sterling wheel will rotate4th stage of tracking loopcauses the system to change, resulting in an output (o(t)) ex: vehicle changes lateral positiontracking loopcommand input... display... human operator... control... systemwhere do errors come from in the tracking loop?-change in "commanded" position of the target (lane shift) -system disturbances (wind gusts may push vehicle for lane center - changes in environmentfundamental task of the operator-keep o(t) = i(t) -in other words, keep e(t) = 0time delay (between input and output of system)-universally harmful to tracking performance -greater costs for greater delays -requires operator to predict future system statesthree sources of delay:-processing time -computational -communicationprocessing timethere will always be some delay in process time, we cannot instantaneously recognize error and decide on the correct response -also known as effective time delay)time needed determined by the complexity of the transfer function:- 0th and 1st order systems are tracked with time delays of ~150-300 ms -2nd order system: 400-500 ms more complex error correction decision requiredexample of computational delayssluggish softwaregainthe ratio of the amplitude of the output to the of the input -O/Itradeoff between effort and precision-high gain systems require low effort -high gain also makes it likely that you "overshoot" the target (and overshoot again during a correction, and overshoot again correcting for the overcorrection)-- instability -often advantage for "middle gain" or intermediate gain systemssystems that have high lag and high gain can result in...system instabilitybandwidth (amount of change)-for random signal, best human can do is 2 corrections per second -similar to serial reaction time experiments, 2.5 decisions per secondfor predictable signals...tracking is far superior0 orderposition control -just control position -you provide an input, and after you remove the input, the cursor goes back to its old position1st orderrate control (rate of the cursor)2nd orderacceleration controlas order increases...manual control becomes more difficultwhich system is best for humans?0th and 1st equivalent2nd order systems should beavoided -operators must perceived higher error derivatives continuously -near constant monitoring reduces cognitive capacity for other tasks -errors and subjective workload increases significantlydirect positiontouchscreens, stylusindirect positionmouseindirect velocityjoystickcategories of input devices:-direction position -indirect position -indirect velocity -voicea signal trial is when...a signal is present -ex: TSA baggage screener encounters a bag and that bag has a knife in itnoise trials are when..no target is present (but random variation within this noise may look like a target sometimes)direct positiontouchscreens, stylus1st order