Rankings and scores are two common data types used by judges to express preferences and/or perceptions of quality in a collection of objects. Numerous models exist to study data of each type separately, but no unified statistical model captures both data types simultaneously without first performing data conversion. We propose the Mallows-Binomial model to close this gap, which combines a Mallows $\phi$ ranking model with Binomial score models through shared parameters that quantify object quality, a consensus ranking, and the level of consensus among judges. We propose an efficient tree-search algorithm to calculate the exact MLE of model parameters, study statistical properties of the model both analytically and through simulation, and apply our model to real data from an instance of grant panel review that collected both scores and partial rankings. Furthermore, we demonstrate how model outputs can be used to rank objects with confidence. The proposed model is shown to sensibly combine information from both scores and rankings to quantify object quality and measure consensus with appropriate levels of statistical uncertainty.
more »
« less
Uncertainty-aware Score Distribution Learning for Action Quality Assessment
Assessing action quality from videos has attracted growing attention in recent years. Most existing approaches usually tackle this problem based on regression algorithms, which ignore the intrinsic ambiguity in the score labels caused by multiple judges or their subjective appraisals. To address this issue, we propose an uncertainty-aware score distribution learning (USDL) approach for action quality assessment (AQA). Specifically, we regard an action as an instance associated with a score distribution, which describes the probability of different evaluated scores. Moreover, under the circumstance where fine-grained score labels are available (e.g., difficulty degree of an action or multiple scores from different judges), we further devise a multi-path uncertainty-aware score distributions learning
(MUSDL) method to explore the disentangled components of a score. We conduct experiments on three AQA datasets containing various Olympic actions and surgical activities, where our approaches set new state-of-the-arts under the Spearman’s Rank Correlation.
more »
« less
- Award ID(s):
- 1815561
- NSF-PAR ID:
- 10161343
- Date Published:
- Journal Name:
- IEEE/CVF Conf. on Computer Vision and Pattern Recognition
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
to inconsistencies between annotators. The low inter-evaluator agreement arises due to the complex nature of emotions. Conventional approaches average scores provided by multiple annotators. While this approach reduces the influence of dissident annotations, previous studies have showed the value of considering individual evaluations to better capture the underlying ground-truth. One of these approaches is the qualitative agreement (QA) method, which provides an alternative framework that captures the inherent trends amongst the annotators. While previous studies have focused on using the QA method for time-continuous annotations from a fixed number of annotators, most emotional databases are annotated with attributes at the sentence-level (e.g., one global score per sentence). This study proposes a novel formulation based on the QA framework to estimate reliable sentence-level annotations for preferencelearning. The proposed relative labels between pairs of sentences capture consistent trends across evaluators. The experimental evaluation shows that preference-learning methods to rank-order emotional attributes trained with the proposed QAbased labels achieve significantly better performance than the same algorithms trained with relative scores obtained by averaging absolute scores across annotators. These results show the benefits of QA-based labels for preference-learning using sentence-level annotations.more » « less
-
more » « less
Abstract Background In many grant review settings, proposals are selected for funding on the basis of summary statistics of review ratings. Challenges of this approach (including the presence of ties and unclear ordering of funding preference for proposals) could be mitigated if rankings such as top-k preferences or paired comparisons, which are local evaluations that enforce ordering across proposals, were also collected and incorporated in the analysis of review ratings. However, analyzing ratings and rankings simultaneously has not been done until recently. This paper describes a practical method for integrating rankings and scores and demonstrates its usefulness for making funding decisions in real-world applications.
Methods We first present the application of our existing joint model for rankings and ratings, the Mallows-Binomial, in obtaining an integrated score for each proposal and generating the induced preference ordering. We then apply this methodology to several theoretical “toy” examples of rating and ranking data, designed to demonstrate specific properties of the model. We then describe an innovative protocol for collecting rankings of the top-six proposals as an add-on to the typical peer review scoring procedures and provide a case study using actual peer review data to exemplify the output and how the model can appropriately resolve judges’ evaluations.
Results For the theoretical examples, we show how the model can provide a preference order to equally rated proposals by incorporating rankings, to proposals using ratings and only partial rankings (and how they differ from a ratings-only approach) and to proposals where judges provide internally inconsistent ratings/rankings and outlier scoring. Finally, we discuss how, using real world panel data, this method can provide information about funding priority with a level of accuracy in a well-suited format for research funding decisions.
Conclusions A methodology is provided to collect and employ both rating and ranking data in peer review assessments of proposal submission quality, highlighting several advantages over methods relying on ratings alone. This method leverages information to most accurately distill reviewer opinion into a useful output to make an informed funding decision and is general enough to be applied to settings such as in the NIH panel review process.
-
Obeid, Iyad ; Selesnick, Ivan ; Picone, Joseph (Ed.)The Temple University Hospital Seizure Detection Corpus (TUSZ) [1] has been in distribution since April 2017. It is a subset of the TUH EEG Corpus (TUEG) [2] and the most frequently requested corpus from our 3,000+ subscribers. It was recently featured as the challenge task in the Neureka 2020 Epilepsy Challenge [3]. A summary of the development of the corpus is shown below in Table 1. The TUSZ Corpus is a fully annotated corpus, which means every seizure event that occurs within its files has been annotated. The data is selected from TUEG using a screening process that identifies files most likely to contain seizures [1]. Approximately 7% of the TUEG data contains a seizure event, so it is important we triage TUEG for high yield data. One hour of EEG data requires approximately one hour of human labor to complete annotation using the pipeline described below, so it is important from a financial standpoint that we accurately triage data. A summary of the labels being used to annotate the data is shown in Table 2. Certain standards are put into place to optimize the annotation process while not sacrificing consistency. Due to the nature of EEG recordings, some records start off with a segment of calibration. This portion of the EEG is instantly recognizable and transitions from what resembles lead artifact to a flat line on all the channels. For the sake of seizure annotation, the calibration is ignored, and no time is wasted on it. During the identification of seizure events, a hard “3 second rule” is used to determine whether two events should be combined into a single larger event. This greatly reduces the time that it takes to annotate a file with multiple events occurring in succession. In addition to the required minimum 3 second gap between seizures, part of our standard dictates that no seizure less than 3 seconds be annotated. Although there is no universally accepted definition for how long a seizure must be, we find that it is difficult to discern with confidence between burst suppression or other morphologically similar impressions when the event is only a couple seconds long. This is due to several reasons, the most notable being the lack of evolution which is oftentimes crucial for the determination of a seizure. After the EEG files have been triaged, a team of annotators at NEDC is provided with the files to begin data annotation. An example of an annotation is shown in Figure 1. A summary of the workflow for our annotation process is shown in Figure 2. Several passes are performed over the data to ensure the annotations are accurate. Each file undergoes three passes to ensure that no seizures were missed or misidentified. The first pass of TUSZ involves identifying which files contain seizures and annotating them using our annotation tool. The time it takes to fully annotate a file can vary drastically depending on the specific characteristics of each file; however, on average a file containing multiple seizures takes 7 minutes to fully annotate. This includes the time that it takes to read the patient report as well as traverse through the entire file. Once an event has been identified, the start and stop time for the seizure is stored in our annotation tool. This is done on a channel by channel basis resulting in an accurate representation of the seizure spreading across different parts of the brain. Files that do not contain any seizures take approximately 3 minutes to complete. Even though there is no annotation being made, the file is still carefully examined to make sure that nothing was overlooked. In addition to solely scrolling through a file from start to finish, a file is often examined through different lenses. Depending on the situation, low pass filters are used, as well as increasing the amplitude of certain channels. These techniques are never used in isolation and are meant to further increase our confidence that nothing was missed. Once each file in a given set has been looked at once, the annotators start the review process. The reviewer checks a file and comments any changes that they recommend. This takes about 3 minutes per seizure containing file, which is significantly less time than the first pass. After each file has been commented on, the third pass commences. This step takes about 5 minutes per seizure file and requires the reviewer to accept or reject the changes that the second reviewer suggested. Since tangible changes are made to the annotation using the annotation tool, this step takes a bit longer than the previous one. Assuming 18% of the files contain seizures, a set of 1,000 files takes roughly 127 work hours to annotate. Before an annotator contributes to the data interpretation pipeline, they are trained for several weeks on previous datasets. A new annotator is able to be trained using data that resembles what they would see under normal circumstances. An additional benefit of using released data to train is that it serves as a means of constantly checking our work. If a trainee stumbles across an event that was not previously annotated, it is promptly added, and the data release is updated. It takes about three months to train an annotator to a point where their annotations can be trusted. Even though we carefully screen potential annotators during the hiring process, only about 25% of the annotators we hire survive more than one year doing this work. To ensure that the annotators are consistent in their annotations, the team conducts an interrater agreement evaluation periodically to ensure that there is a consensus within the team. The annotation standards are discussed in Ochal et al. [4]. An extended discussion of interrater agreement can be found in Shah et al. [5]. The most recent release of TUSZ, v1.5.2, represents our efforts to review the quality of the annotations for two upcoming challenges we hosted: an internal deep learning challenge at IBM [6] and the Neureka 2020 Epilepsy Challenge [3]. One of the biggest changes that was made to the annotations was the imposition of a stricter standard for determining the start and stop time of a seizure. Although evolution is still included in the annotations, the start times were altered to start when the spike-wave pattern becomes distinct as opposed to merely when the signal starts to shift from background. This cuts down on background that was mislabeled as a seizure. For seizure end times, all post ictal slowing that was included was removed. The recent release of v1.5.2 did not include any additional data files. Two EEG files had been added because, originally, they were corrupted in v1.5.1 but were able to be retrieved and added for the latest release. The progression from v1.5.0 to v1.5.1 and later to v1.5.2, included the re-annotation of all of the EEG files in order to develop a confident dataset regarding seizure identification. Starting with v1.4.0, we have also developed a blind evaluation set that is withheld for use in competitions. The annotation team is currently working on the next release for TUSZ, v1.6.0, which is expected to occur in August 2020. It will include new data from 2016 to mid-2019. This release will contain 2,296 files from 2016 as well as several thousand files representing the remaining data through mid-2019. In addition to files that were obtained with our standard triaging process, a part of this release consists of EEG files that do not have associated patient reports. Since actual seizure events are in short supply, we are mining a large chunk of data for which we have EEG recordings but no reports. Some of this data contains interesting seizure events collected during long-term EEG sessions or data collected from patients with a history of frequent seizures. It is being mined to increase the number of files in the corpus that have at least one seizure event. We expect v1.6.0 to be released before IEEE SPMB 2020. The TUAR Corpus is an open-source database that is currently available for use by any registered member of our consortium. To register and receive access, please follow the instructions provided at this web page: https://www.isip.piconepress.com/projects/tuh_eeg/html/downloads.shtml. The data is located here: https://www.isip.piconepress.com/projects/tuh_eeg/downloads/tuh_eeg_artifact/v2.0.0/.more » « less
-
Abstract Expert testimony varies in scientific quality and jurors have a difficult time evaluating evidence quality (McAuliff et al., 2009). In the current study, we apply Fuzzy Trace Theory principles, examining whether visual and gist aids help jurors calibrate to the strength of scientific evidence. Additionally we were interested in the role of jurors’ individual differences in scientific reasoning skills in their understanding of case evidence. Contrary to our preregistered hypotheses, there was no effect of evidence condition or gist aid on evidence understanding. However, individual differences between jurors’ numeracy skills predicted evidence understanding. Summary Poor-quality expert evidence is sometimes admitted into court (Smithburn, 2004). Jurors’ calibration to evidence strength varies widely and is not robustly understood. For instance, previous research has established jurors lack understanding of the role of control groups, confounds, and sample sizes in scientific research (McAuliff, Kovera, & Nunez, 2009; Mill, Gray, & Mandel, 1994). Still others have found that jurors can distinguish weak from strong evidence when the evidence is presented alone, yet not when simultaneously presented with case details (Smith, Bull, & Holliday, 2011). This research highlights the need to present evidence to jurors in a way they can understand. Fuzzy Trace Theory purports that people encode information in exact, verbatim representations and through “gist” representations, which represent summary of meaning (Reyna & Brainerd, 1995). It is possible that the presenting complex scientific evidence to people with verbatim content or appealing to the gist, or bottom-line meaning of the information may influence juror understanding of that evidence. Application of Fuzzy Trace Theory in the medical field has shown that gist representations are beneficial for helping laypeople better understand risk and benefits of medical treatment (Brust-Renck, Reyna, Wilhelms, & Lazar, 2016). Yet, little research has applied Fuzzy Trace Theory to information comprehension and application within the context of a jury (c.f. Reyna et. al., 2015). Additionally, it is likely that jurors’ individual characteristics, such as scientific reasoning abilities and cognitive tendencies, influence their ability to understand and apply complex scientific information (Coutinho, 2006). Methods The purpose of this study was to examine how jurors calibrate to the strength of scientific information, and whether individual difference variables and gist aids inspired by Fuzzy Trace Theory help jurors better understand complicated science of differing quality. We used a 2 (quality of scientific evidence: high vs. low) x 2 (decision aid to improve calibration - gist information vs. no gist information), between-subjects design. All hypotheses were preregistered on the Open Science Framework. Jury-eligible community participants (430 jurors across 90 juries; Mage = 37.58, SD = 16.17, 58% female, 56.93% White). Each jury was randomly assigned to one of the four possible conditions. Participants were asked to individually fill out measures related to their scientific reasoning skills prior to watching a mock jury trial. The trial was about an armed bank robbery and consisted of various pieces of testimony and evidence (e.g. an eyewitness testimony, police lineup identification, and a sweatshirt found with the stolen bank money). The key piece of evidence was mitochondrial DNA (mtDNA) evidence collected from hair on a sweatshirt (materials from Hans et al., 2011). Two experts presented opposing opinions about the scientific evidence related to the mtDNA match estimate for the defendant’s identification. The quality and content of this mtDNA evidence differed based on the two conditions. The high quality evidence condition used a larger database than the low quality evidence to compare to the mtDNA sample and could exclude a larger percentage of people. In the decision aid condition, experts in the gist information group presented gist aid inspired visuals and examples to help explain the proportion of people that could not be excluded as a match. Those in the no gist information group were not given any aid to help them understand the mtDNA evidence presented. After viewing the trial, participants filled out a questionnaire on how well they understood the mtDNA evidence and their overall judgments of the case (e.g. verdict, witness credibility, scientific evidence strength). They filled this questionnaire out again after a 45-minute deliberation. Measures We measured Attitudes Toward Science (ATS) with indices of scientific promise and scientific reservations (Hans et al., 2011; originally developed by National Science Board, 2004; 2006). We used Drummond and Fischhoff’s (2015) Scientific Reasoning Scale (SRS) to measure scientific reasoning skills. Weller et al.’s (2012) Numeracy Scale (WNS) measured proficiency in reasoning with quantitative information. The NFC-Short Form (Cacioppo et al., 1984) measured need for cognition. We developed a 20-item multiple-choice comprehension test for the mtDNA scientific information in the cases (modeled on Hans et al., 2011, and McAuliff et al., 2009). Participants were shown 20 statements related to DNA evidence and asked whether these statements were True or False. The test was then scored out of 20 points. Results For this project, we measured calibration to the scientific evidence in a few different ways. We are building a full model with these various operationalizations to be presented at APLS, but focus only on one of the calibration DVs (i.e., objective understanding of the mtDNA evidence) in the current proposal. We conducted a general linear model with total score on the mtDNA understanding measure as the DV and quality of scientific evidence condition, decision aid condition, and the four individual difference measures (i.e., NFC, ATS, WNS, and SRS) as predictors. Contrary to our main hypotheses, neither evidence quality nor decision aid condition affected juror understanding. However, the individual difference variables did: we found significant main effects for Scientific Reasoning Skills, F(1, 427) = 16.03, p <.001, np2 = .04, Weller Numeracy Scale, F(1, 427) = 15.19, p <.001, np2 = .03, and Need for Cognition, F(1, 427) = 16.80, p <.001, np2 = .04, such that those who scored higher on these measures displayed better understanding of the scientific evidence. In addition there was a significant interaction of evidence quality condition and scores on the Weller’s Numeracy Scale, F(1, 427) = 4.10, p = .04, np2 = .01. Further results will be discussed. Discussion These data suggest jurors are not sensitive to differences in the quality of scientific mtDNA evidence, and also that our attempt at helping sensitize them with Fuzzy Trace Theory-inspired aids did not improve calibration. Individual scientific reasoning abilities and general cognition styles were better predictors of understanding this scientific information. These results suggest a need for further exploration of approaches to help jurors differentiate between high and low quality evidence. Note: The 3rd author was supported by an AP-LS AP Award for her role in this research. Learning Objective: Participants will be able to describe how individual differences in scientific reasoning skills help jurors understand complex scientific evidence.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
