4323513 - Grade: 8.3
Man-Machine-Systems (ME41080)
Technische Universiteit Delft
Aanbevolen voor jou
Reacties
Preview tekst
Question 1A Sheridan classified the study of man-machine-systems chronologically in 3 phases: 1) knobs and dial, 2) borrowed engineering models, and 3)human-computer interaction. [1] 1. Knobs and Dials. This phase is focused on the design of the computer-human interface. The use of a computer system requires an interface to interpret the information given by the computer. Research in this phase investigated ‘how to design’ interface for which the information transition to the user is optimized. [2] This research is NOT applied in the article of Falcone et al. (2012). 2. Borrowed Engineering Models. Through the use of borrowed engineering models researchers could study and predict human-system performance. [1] Examples of borrowed engineering models are signal detection theory, feedback control theory an information theory. Falcone et al. (2012) assessed in his research whether the use of tCDS can enhance perceptual sensitivity (d’) in detection tasks. In this research the authors used the signal detection theory in order to evaluate their hypothesis: applying tCDS would improve d’, but tCDS would not affect the response bias (). The signal detection theory is able to distinguish the contribution of d’ and [3]Based on these findings the article belongs to the borrowed engineering model phase. 1. Human- Computer Interaction. Human-Computer interaction reflects a phase wherein manual systems are transitioned to automated systems, where the human only has a supervisory role. In automated systems the stimuli or signals are detected by the computer, but the human should detect the automation failures. [4] In the article of Falcone et al.(2012) the researchers used computer-generated virtual reality environments to examine how subjects detect threats. [3] The virtual reality(VR) images are generated by a computer and the VR tracks the user in real time. [5] Furthermore the location for the application of tCDS was determined by a fMRI. Virtual reality and this type of neuroimaging belong to the human-computer interaction phase. Therefore a part of this research belongs to the human-computer interaction phase. Question 1B By the means of Matlab the figure is created that show the PDF of the noise and the signal. In the Matlab figure the correct rejections, correct hits, false positives, false negatives, d’ and β are simulated (can be seen in appendix A). For the computation of the code of Matlab the equations of the signal detection theory (z-scores, perceptual sensitivity, criterion value and response bias) are used from the book of MacMillan and lecture 4 of the course Man-Machine-System. [6] [7] The z-scores are computed by using the norminv Matlab function of the hit rate and the false alarm rate. In figure 1 the equations are shown. 𝑑′ = 𝑧(𝐻) − 𝑧(𝐹) 𝐶 = −𝑧(𝐻) 𝛽 = 𝑒 𝑑′×((𝐶−𝑑′)/2) Figure 1: Equations used in the Matlab code to compute the probability density functions [6] [7] Perceptual sensitivity explains the detectability of a signal: how hard is it to detect the information bearing stimulus from other events. A high d’ indicates that the signal is easily detected in compared to a low d’. The signal detection theory states that d’ is unaffected by the response bias. The response bias explains to what extent the subject is inclined to give a certain response, i. the willingness to say that the subject detected the signal. For example the probability of the subject to answer yes to a stimuli is 0, then the observers is inclined to say yes more often than no. When β < 1, a liberal bias exist: the subject has the tendency to say yes. If the subjects have no favour towards responding yes or no, β = 1. [8] [9] The outcomes from the Matlab script are in agreement values from the article, see table 1 for the display of d’ and β. A slight conservative response bias can be assessed from this result. The perceptual sensitivity is quite high, corresponding with a relative high hit rate and low false alarm rate from the article. Table 1: perceptual sensitivity and response bias of training 4 from the Matlab script β d’ Training 4 - 2 mA 1 1 Question 1C The vigilance decrement, i. a drop of hit rate as a function of time can be explained on the basis of the arousal theory of vigilance. Studies have shown that the hit rate is significantly higher if the event rate is higher. [10] This result coincides with the arousal theory: if the subject is exposed to more stimuli it will experience a more active psychological state, resulting in a higher hit rate. Furthermore studies have shown that providing people with neuro stimulants (caffeine) increases their hit rate. This phenomenon affirms the arousal theory. Additionally other studies have shown that other factors exit which influence the vigilance decrement. According to Warm et al.(2008) vigilance tasks have a high mental workload and induce stress and frustration over time. [11] These factors decrease the perceptual sensitivity. [6] Slazma (1999) concluded that stress and frustration increase over time during a vigilance task at the hand of self-reporting questionnaire. [12] When I reflect this on my own DSSQ test scores from report 1 question 2A, I identify an increase in worry and decrease in distress. Thus my scores are in agreement with the theory. By means of the Matlab file theory can be investigated. I decrease the hit rate to 26%. This results in a decrease of the perceptual sensitivity, (d’= 0). The response bias becomes slightly more conservative, β= 1. Therefore can be concluded that a decrease in hit rate can be attributed to the detectability of the signal, d’. Question 1D The study design of Faclone et al.(2012) was an experimental trial. In the article it is mentioned that the subjects were assigned randomly to, the ‘sham’ ore active test group. The purpose of randomization is to minimize the differences between the two groups in the trial. But in this article it is not clear whether randomization was successful or even properly applied. Furthermore in order to assess homogeneity in the groups a patient characteristics table should be implemented in this article. In the selection process the researchers only used exclusion criteria, to realize more homogenous groups, the authors could have implemented inclusion criteria. This experimental design was a single blinded design, in order to prevent observer bias a double blinded experiment is desired. I also think that the authors make strong claims about the possibilities of tCDS. The results of the trial are positive in favour of the test group for the detection task. However the question that needs to be asked and answered is whether these result are relevant and how they can be applied in real life situations. The author argues that tCDS can be used to enhance learning. But to wat degree? Coffman et al. (2014) demonstrated that tCDS also augmented three other fundamental cognitive functions of the exposed subject: attention, learning and memory process. [13] In current literature it is hypothesized that enhancement of the functions can lead to augmentation of higher order processes, such as decision-making and problem solving. [14] If this hypothesis is true, it will have a great social consequence: Clark (2014) stated that “cognitive enhancement might lead to an arms race, where all people are required to use enhancement in order to stay competitive”. [14] [15] This will result in inequality issue between the people who can afford tCDS and people who cannot. Therefore an important ethical question needs to be asked: “who should be able to use brain stimulation”. [16] Question 2A In figure 29 of the book of Fitss and Posner (1967) the reaction is plotted as a function of the amount of information in the stimulus series (expressed as average amount of information in bitts) or otherwise known as the stimulus uncertainty. This figure is a result from the research of Hyman (1952). [17] The study of Hyman investigated the relationship between different the amount of information per stimuli and the reaction time. The complexity of a task can be increased by modifying different aspects of the given stimuli: number of possible stimuli, the frequency of the stimuli and stimuli intensity. For instance if the task requires the subject to discriminate between the stimuli and to make a choice how to response, the reaction time will increase. Hick (1952) assessed that response time was proportional to the log of the number of different possible stimuli. Furthermore when the frequency of the stimuli will increase, the reaction time will decrease. [17] Hyman investigated his hypothesis by varying the amount of information per stimulus signal through three different experiments: 1) varying the number of alternatives (which were equally probable), 2)changing the probabilities of occurrence of the alternatives and the number of alternatives, 3) varying the number of alternatives and introducing sequential interdependencies between the occurring stimuli and the successive stimuli. [18] This figures confirms the hypothesis of Hyman(1952): the results of the research suggest that “the reaction time can be considered a linear function of stimulus information”, because the three experiments have approximately the same linear fit. Due to the linear relationship when the average amount of information, i. the bitts, increase the reaction time will also increase. [18] This theory was reviewed and reaffirmed by Klapp (2010). Question 3A Saccade is a sudden, reflexive, simultaneous and small eye movement which occurs when the point of fixation of the fovea centralis needs to be changed in order to acquire 100 % visual acuity of the object that needs to be visualized. The focus of visual attention is in correspondence with the focus of cognitive attention. For example when you are reading, your eyes make a jump from each word in order to align the focus point with the fovea in order to interpret the data. This is called a fixation. Fixation occurs when the eyes remain almost stationary in order to process the visual information. Therefore can be concluded that saccades facilitate the movement which is required for fixation. The saccade has a duration of 10-40 ms. [22] The duration of fixation is approximately between 100 and 500 ms. During the saccade the visual information is not processed by the brain, because a saccade would be perceived as a ‘blur’ and should not be confused with actual visual information. [23] In literature exist different calculation methods to analyze the saccades and fixations: velocity-based algorithms, dispersion-based algorithms and area-based algorithms. [24] The data provided in de excel file are x and y coordinates of the focus position on the screen. In accordance with the provided date the best method for analyzation is the Velocity-Based I-VT method. This method separates the saccades and based on point-to-point velocities, “each velocity is computed as the distance between the current point and the next point, I-VT then classifies the points based on a simple threshold”. [24] [25] I used the pseudocode from Salvucci et al(2010) to construct my algorithm (in appendix B). The threshold used for saccade is >2000 pixels/second, this is from the article of Eisma(2017). Before the data is used to measure the angular distance a Savatzky-Golay filter is applied as suggested by the article of Nystrom et al and the blink is removed. (2010). Question 3B I provided 4 three figures in the appendix to visualize my data in appendix C. The first figure is the path of the x and y coordinates. The second figure shows the measured velocities and the threshold (2000 pixels/second). Velocities that cross the threshold are considered as saccades. The third figure correlates the video to the displacement in the x-direction. The red car from the right is seen at t=13s the video. In figure 3 you see at t=13s for ~1 sec no x-displacement is measured and in figure 2 you see around t=13s a long fixation. This fixation is followed by a saccade, this is due the car coming from the left at t=14s. This saccade followed again by a fixation. In figure 3 at t=14s a large displacement (in x-direction) is seen. Due to the incoming car in the left, the visual attentions moves to the other side of the screen. The fourth figure shows the y-displacement. In this figure in general no big displacement is seen. Therefore I conclude that the visual attention is focused on the midst of the video. The midst of the video shows the road, so the observer keeps his eyes on the road. The mean fixation time is 302 ms, this is in correspondence with the values from literature (100-500 ms). The average saccade amplitude is 11049 pixels/sec. Converted to degrees/sec this is ~186 degrees/sec. Literature sets a threshold <100 degrees/sec for fixations and >300 degrees for saccades. [24] [27] The [degrees/sec] unit is not in complete agreement with the literature. Question 3C The human being cannot see everything, the human brain has a limited capacity to visual process information. Itti et al (2000) stated that “Only a small fraction of the information is registered by the visual system”. [28] Therefore we must discriminate in where we focus our visual attention. The SEEV model, presented by Wickens (2008) predicts the probability that a given event will attract visual attention based on 4 variables: salience, effort, expectancy and value. [29] The eye-tracking data can be explained according the SEEV model. There are multiple sources (6) that influence the visual information access. The most dominant one is salience. Salience describes the distinctness of a signal from other signals and therefore the likelihood to capture attention If a signal has a high salience your visual attention will probably focus on the given signal. Salience is a bottom up controlled by the attention capturing property. The reaction on salient signals are sensory driven cues and is a primary unconscious reaction. [27] Therefore most warnings are salient, because it requires direct visual attention. Other examples of are salient signals are bright flashes or sounds, sudden moving images. [29]The salience of a signal or object can be influenced by the colour, contrast and size. In the video the at t=13s, when the red car comes into the observer’s view. This can be judged as a salient observation due to the red colour and the movement towards the crossroad. Expectancy is influenced by the bandwidth(event rate) and signalling (prior context). Due to a higher event rate the observer will expect more events at that specific area of interest (AOI), resulting in more visual attention on that AOI. The observer sees in the video an crossroad. People who are experienced in traffic expect that automobiles can approach the crossroad from left or right. This corresponds with figure 3, because the visual attention shifts from the left side of the video to the right side. The observer expect that an event will occur at this AOI, therefore the visual attention is focused on these AOI’s. In order the have a safe (road)trip, the driver has to keep his eyes on the road. This source is where the driver acquires the most valuable information which are necessary to execute actions and make decision for driving. In figure 4 can be seen that this is in correspondence with the above theory, i. the visual attention of the observer is on the road. Appendix C Appendix C: Figure 1 of 3B eye-tracking path Appendix C: Figure 2 of 3B visualization of the saccades and the fixations Appendix C: Figure 3 of 3B: x-direction displacement Appendix C: Figure 4 of 3B: y-direction displacement Appendix C: Outcomes from the matlabscript [16] V. Q. Walsh, “Ethics and Social Risks in Brain Stimulation,” Brain stimulation, vol. 6, pp. 715-717, 2013. [17] P. M. Fits and M. I. Posner, “Human Performance,” California, Cole Publishing Company, 1967, pp. 90-101. [18] R. Hyman, “Stimulus information as a determinant of reaction time,” in PhD thesis at the john Hopkins Unversity Faculty of Philosophy, 1952. [19] A. Jain, R. Bansal, A. Kumar and K. Singh, “A comparative study of visual and auditory reaction times on the basis of gender and physical activity levels of medical first year students,” International Journal of Applied and Basic Medical Research, no. May 2, pp. 124-127, 2015. [20] P. M. Fitts and C. M. Seeger, “SR compatibility: spatial characteristics of stimulus and response codes,” Journal of Experimental Psychology, no. 46, pp. 199-210, 1953. [21] B. J. Kemp, “Reaction time of young and elderly subjects in relation to perceptual deprivation and signal-on versus signal-off condition,” Developmental Psychology, vol. 8, pp. 268-272, 1973. [22] D. Purves, G. Augustine, D. Fitzpatrick and e. al, “Ch Types of Eye Movements and Their Functions,” in Neuroscience, Sunderland, Sinauer Associates, 2001. [23] K. B. Ibbotson M, “Visual Perception and Saccadic Eye Movements.,” Current Opinion in Neurobiology, 21(4), p. 553–558., 2001. [24] D. D. Salvucci and J. H. Goldberg, “Identifying fixations and saccades in eye-tracking protocols. In Proceedings of the Eye Tracking Research and Applications,” Symposium New York: ACM Press, pp. 71-78, 2000. [25] C. J. Erkelens and I. M. L. C. Vogels, “The initial direction and landing position of saccades. Eye movement research: mechanisms, processes and applications,” New york: Elsevier, pp. 133144, 1995. [26] “User manual: EyeLink 1000® Data Viewer”. [27] F. M. Marchak, “Analysis of Eye Movement Data”. [28] L. Itti and C. Koch, “ A saliency-based search mechanism for overt and covert shifts of visual attention,” Vision Research, vol. 40, no. 10-12, pp. 1489-15106, 2000. [29] C. D. Wickens and J. S. McCarley, “Applied attention theory,” in Visual attention control, scanning, and information sampling, Boca Raton, CRC Press., 2008, p. 41–61.
4323513 - Grade: 8.3
Vak: Man-Machine-Systems (ME41080)
Universiteit: Technische Universiteit Delft
- Ontdek meer van: