Alexia Zoumpoulaki

A multi-agent simulation framework for emergency evacuations incorporating personality and emotions

Software simulations of building evacuation during emergency can provide rich qualitative and quantitative results for safety analysis However, the majority of them do not take into account current surveys on human behaviors under stressful situations that explain the important role of personality and emotions in crowd behaviors during evacuations In this paper we propose a framework for designing evacuation simulations that is based on a multi-agent BDI architecture enhanced with the OCEAN model of personality and the OCC model of emotions.

Data-driven region-of-interest selection without inflating Type I error rate.

In ERP and other large multidimensional neuroscience data sets, researchers often select regions of interest (ROIs) for analysis. The method of ROI selection can critically affect the conclusions of a study by causing the researcher to miss effects in the data or to detect spurious effects. In practice, to avoid inflating Type I error rate (i.e., false positives), ROIs are often based on a priori hypotheses or independent information. However, this can be insensitive to experiment-specific variations in effect location (e.g., latency shifts) reducing power to detect effects. Data-driven ROI selection, in contrast, is nonindependent and uses the data under analysis to determine ROI positions. Therefore, it has potential to select ROIs based on experiment-specific information and increase power for detecting effects. However, data-driven methods have been criticized because they can substantially inflate Type I error rate. Here, we demonstrate, using simulations of simple ERP experiments, that data-driven ROI selection can indeed be more powerful than a priori hypotheses or independent information. Furthermore, we show that data-driven ROI selection using the aggregate grand average from trials (AGAT), despite being based on the data at hand, can be safely used for ROI selection under many circumstances. However, when there is a noise difference between conditions, using the AGAT can inflate Type I error and should be avoided. We identify critical assumptions for use of the AGAT and provide a basis for researchers to use, and reviewers to assess, data-driven methods of ROI localization in ERP and other studies.

Resampling the peak, some dos and don'ts

Resampling techniques are used widely within the ERP community to assess statistical significance and especially in the deception detection literature. Here, we argue that because of statistical bias, bootstrap should not be used in combination with methods like peak-to-peak. Instead, permutation tests provide a more appropriate alternative.

Identification of memory reactivation during sleep by EEG classification

ABSTRACT Memory reactivation during sleep is critical for consolidation, but also extremely difficult to measure as it is subtle, distributed and temporally unpredictable. This article reports a novel method for detecting such reactivation in standard sleep recordings. During learning, participants produced a complex sequence of finger presses, with each finger cued by a distinct audio‐visual stimulus. Auditory cues were then re‐played during subsequent sleep to trigger neural reactivation through a method known as targeted memory reactivation (TMR). Next, we used electroencephalography data from the learning session to train a machine learning classifier, and then applied this classifier to sleep data to determine how successfully each tone had elicited memory reactivation. Neural reactivation was classified above chance in all participants when TMR was applied in SWS, and in 5 of the 14 participants to whom TMR was applied in N2. Classification success reduced across numerous repetitions of the tone cue, suggesting either a gradually reducing responsiveness to such cues or a plasticity‐related change in the neural signature as a result of cueing. We believe this method will be valuable for future investigations of memory consolidation. HIGHLIGHTSWe developed a method for detecting neural reactivation during sleep using EEG classifiers.We detected cued memory reactivation during SWS, and to a lesser extent during S2.Classification decreased as more cues were presented, suggesting saturation.Discreet wavelet features were more useful than EEG or spectral features.

A new method for detecting deception in Event Related Potentials using individual-specific weight templates

A new method called the weight template (WT) is proposed for classifying Event related potentials (ERPs) into deceiving and non-deceiving. In this study, EEG data from two P300-based lie detection experiments were analyzed to demonstrate the efficiency of the WT method in detecting deception. A comparison was made with a common method used to measure P300 presence, called Peak-to-Peak, which is believed to be more accurate than other methods in measuring P300 amplitudes [1, 2]. One experiment consisted of presenting participants with birth date stimuli and 12 participants were instructed to lie about their own birthday. The other experiment consisted of 15 participants who were instructed to lie about their first names [3]. Using simulated EEG data [4], Receiver Operating Characteristic (ROC) curves were also generated to examine the efficiency of the proposed method in detecting deception in low signal-to-noise ERPs. Typically, P300-based lie detection systems employ the P300 component to detect concealed information. They present three stimulus types: Probes (P), which represent concealed information or crime details and can be recognized only by the guilty person; Irrelevants (I), which are frequent and task (crime)-irrelevant, and Targets (T), which are irrelevant items, but participants are asked to do a task whenever they see a Target. For practical lie detection, the key comparison is between Probe and Irrelevant ERPs, since, for the nondeceiver, the former would be an Irrelevant. Importantly, the Probe for a deceiver typically generates a P300 ERP component, which is absent for the Irrelevant. The principle underlying the WT method is that as the Target stimulus is task-relevant, it will evoke a robust P300 pattern for each subject, which we hypothesize is characteristic in form and polarity of that individuals P300. Accordingly, this TERP can serve as an individual-specific template, with which to search for the Probe P300. Specifically, the difference between Tand IERPs was used as a template (i.e. effectively as a kernel) and this template was applied to Pand IERPs. Using such a template, with some pre-processing steps, we found that the WT achieved significantly better detection performance in comparison to Peak-to-Peak. In the names lie detection, the WT was able to detect deception for 93% in the guilty group compared with 80% by Peak-to peak. The false alarm rates using WT and Peak-to-Peak were 2% and 8% respectively. In the birthdays lie detection, hit rates were 50% using WT and 33% using Peak-to-Peak. The false alarm rates of both methods were 5%. ROC curve analysis also showed that in ERPs with high signal-to-noise ratio (SNR), both methods could detect deception successfully and almost equally. However, the WT performed better in ERPs with low SNR. We thus conclude that the WT is simple and very effective for detecting deception, even in ERPs with low SNR.

ERP latency contrasts using Dynamic Time Warping algorithm

Latency contrasts are central to Event Related Potential (ERP) research. For example, they are used to determine the order and length of cognitive processes or to evaluate how experimental conditions influence processing time. The most popular methods employed by researchers are peak latency, fractional peak and fractional area. However there are difficulties with these methods, which often include acute sensitivity to noise and window placements [1]. In addition, they require parameter settings that are often difficult to justify. In order to address these issues, we propose Dynamic Time Warping (DTW), an algorithm used for measuring similarity between two sequences, and more precisely the use of the warping path and its relationship to the main diagonal as a measure of latency difference. We tested the performance of DTW by comparing it to 25% - 50% fractional area, peak and 50% fractional peak. In addition to the default DTW we also tested the type IIa step pattern, which constraints the resulting warping path [2]. Data were obtained from a deception detection experiment [3] and ERPs were generated from two channels for one condition. Then the second condition was created in two ways: firstly by offsetting the first condition by 100 time points (0.05 ms), and secondly by offsetting the first condition by an amount sampled from a normal distribution with a mean of 100 time points, while also varying the amplitude. In this way, we simulated latency jitter and amplitude variability between conditions as found in real ERP experiments. Each method was applied to windows placed accordingly to relevant experiments. This was done for each one of the channels (P3a at Fz and P3b at Pz). Then noise was added at the power spectrum of human EEG [4], and the performance of each method was evaluated through permutation tests (100 p-values) for different Signal-to-Noise Ratios (SNR). In order to test the methods independently of response bias, we performed ROC (Receiver Operating Characteristic) analysis, where the false positive rate was obtained by generating the second condition without any latency difference. We also compared DTWs sensitivity to window placement against 25% and 50% fractional area, by selecting a fixed time point as the start of the bounding window and then sequentially adjusting the end of the window and calculating the proportion of windows, for which each of the methods failed to determine the correct latency difference. Our analysis shows that the basic DTW performs at the same level as the fractional area method, which outperforms peak and fractional peak. The typeIIa DTW performs better than all the other methods for most SNRs. Although there is a slight inflation of false positives for high SNRs, the ROC analysis shows that it does not substantially affect DTWs performance. At the same time, DTW shows significantly less sensitivity to window placement than fractional area. These results indicate that DTW is a promising technique for determining latency differences, being more robust to noise and window placement, without being subject to the same number of assumptions or parameterisation as the other methods evaluated.