Evaluating model robustness is critical when developing trustworthy models not only to gain deeper understanding of model behavior, strengths, and weaknesses, but also to develop future models that are generalizable and robust across expected environments a model may encounter in deployment. In this paper we present a framework for measuring model robustness for an important but difficult text classification task - deceptive news detection. We evaluate model robustness to out-of-domain data, modality-specific features, and languages other than English. Our investigation focuses on three type of models: LSTM models trained on multiple datasets(Cross-Domain), several fusion LSTM models trained with images and text and evaluated with three state-of-the-art embeddings, BERT ELMo, and GloVe (Cross-Modality), and character-level CNN models trained on multiple languages (Cross-Language). Our analyses reveal a significant drop in performance when testing neural models on out-of-domain data and non-English languages that may be mitigated using diverse training data. We find that with additional image content as input, ELMo embeddings yield significantly fewer errors compared to BERT orGLoVe. Most importantly, this work not only carefully analyzes deception model robustness but also provides a framework of these analyses that can be applied to new models or extended datasets in the future.
With the increasing use of machine-learning driven algorithmic judgements, it is critical to develop models that are robust to evolving or manipulated inputs. We propose an extensive analysis of model robustness against linguistic variation in the setting of deceptive news detection, an important task in the context of misinformation spread online. We consider two prediction tasks and compare three state-of-the-art embeddings to highlight consistent trends in model performance, high confidence misclassifications, and high impact failures. By measuring the effectiveness of adversarial defense strategies and evaluating model susceptibility to adversarial attacks using character- and word-perturbed text, we find that character or mixed ensemble models are the most effective defenses and that character perturbation-based attack tactics are more successful.
Deception detection is an important task that has been a hot research topic due to its potential applications. It can be applied in many areas, from national security (e.g., airport security, jurisprudence, and law enforcement) to real-life applications (e.g., business and computer vision). However, some critical problems still exist and are worth more investigation. One of the significant challenges in the deception detection tasks is the data scarcity problem. Until now, only one multi-modal benchmark open dataset for human deception detection has been released, which contains 121 video clips for deception detection (i.e., 61 for deceptive class and 60 for truthful class). Such an amount of data is hard to drive deep neural network-based methods. Hence, those existing models often suffer from overfitting problems and low generalization ability. Moreover, the ground truth data contains some unusable frames for many factors. However, most of the literature did not pay attention to these problems. Therefore, in this paper, we design a series of data preprocessing methods to deal with the aforementioned problem first. Then, we propose a multi-modal deception detection framework to construct our novel emotional state-based feature and use the open toolkit openSMILE to extract the features from the audio modality. We also design a voting scheme to combine the emotional states information obtained from visual and audio modalities. Finally, we can determine the novel emotion state transformation feature with our self-designed algorithms. In the experiment, we conduct the critical analysis and comparison of the proposed methods with the state-of-the-art multi-modal deception detection methods. The experimental results show that the overall performance of multi-modal deception detection has a significant improvement in the accuracy from 87.77% to 92.78% and the ROC-AUC from 0.9221 to 0.9265.
Automatically identifying fake news from the Internet is a challenging problem in deception detection tasks. Online news is modified constantly during its propagation, e.g., malicious users distort the original truth and make up fake news. However, the continuous evolution process would generate unprecedented fake news and cheat the original model. We present the Fake News Evolution (FNE) dataset: a new dataset tracking the fake news evolution process. Our dataset is composed of 950 paired data, each of which consists of articles representing the three significant phases of the evolution process, which are the truth, the fake news, and the evolved fake news. We observe the features during the evolution and they are the disinformation techniques, text similarity, top 10 keywords, classification accuracy, parts of speech, and sentiment properties.
Automated systems that detect deception in high-stakes situations can enhance societal well-being across medical, social work, and legal domains. Existing models for detecting high-stakes deception in videos have been supervised, but labeled datasets to train models can rarely be collected for most real-world applications. To address this problem, we propose the first multimodal unsupervised transfer learning approach that detects real-world, high-stakes deception in videos without using high-stakes labels. Our subspace-alignment (SA) approach adapts audio-visual representations of deception in lab-controlled low-stakes scenarios to detect deception in real-world, high-stakes situations. Our best unsupervised SA models outperform models without SA, outperform human ability, and perform comparably to a number of existing supervised models. Our research demonstrates the potential for introducing subspace-based transfer learning to model high-stakes deception and other social behaviors in real-world contexts with a scarcity of labeled behavioral data.
Feature extraction is an important process of machine learning and deep learning, as the process make algorithms function more efficiently, and also accurate. In natural language processing used in deception detection such as fake news detection, several ways of feature extraction in statistical aspect had been introduced (e.g. N-gram). In this research, it will be shown that by using deep learning algorithms and alphabet frequencies of the original text of a news without any information about the sequence of the alphabet can actually be used to classify fake news and trustworthy ones in high accuracy (85\%). As this pre-processing method makes the data notably compact but also include the feature that is needed for the classifier, it seems that alphabet frequencies contains some useful features for understanding complex context or meaning of the original text.
Automated deception detection systems can enhance health, justice, and security in society by helping humans detect deceivers in high-stakes situations across medical and legal domains, among others. This paper presents a novel analysis of the discriminative power of dimensional representations of facial affect for automated deception detection, along with interpretable features from visual, vocal, and verbal modalities. We used a video dataset of people communicating truthfully or deceptively in real-world, high-stakes courtroom situations. We leveraged recent advances in automated emotion recognition in-the-wild by implementing a state-of-the-art deep neural network trained on the Aff-Wild database to extract continuous representations of facial valence and facial arousal from speakers. We experimented with unimodal Support Vector Machines (SVM) and SVM-based multimodal fusion methods to identify effective features, modalities, and modeling approaches for detecting deception. Unimodal models trained on facial affect achieved an AUC of 80%, and facial affect contributed towards the highest-performing multimodal approach (adaptive boosting) that achieved an AUC of 91% when tested on speakers who were not part of training sets. This approach achieved a higher AUC than existing automated machine learning approaches that used interpretable visual, vocal, and verbal features to detect deception in this dataset, but did not use facial affect. Across all videos, deceptive and truthful speakers exhibited significant differences in facial valence and facial arousal, contributing computational support to existing psychological theories on affect and deception. The demonstrated importance of facial affect in our models informs and motivates the future development of automated, affect-aware machine learning approaches for modeling and detecting deception and other social behaviors in-the-wild.
Background: Deception detection is a prevalent problem for security practitioners. With a need for more large-scale approaches, automated methods using machine learning have gained traction. However, detection performance still implies considerable error rates. Findings from other domains suggest that hybrid human-machine integrations could offer a viable path in deception detection tasks. Method: We collected a corpus of truthful and deceptive answers about participants' autobiographical intentions (n=1640) and tested whether a combination of supervised machine learning and human judgment could improve deception detection accuracy. Human judges were presented with the outcome of the automated credibility judgment of truthful and deceptive statements. They could either fully overrule it (hybrid-overrule condition) or adjust it within a given boundary (hybrid-adjust condition). Results: The data suggest that in neither of the hybrid conditions did the human judgment add a meaningful contribution. Machine learning in isolation identified truth-tellers and liars with an overall accuracy of 69%. Human involvement through hybrid-overrule decisions brought the accuracy back to the chance level. The hybrid-adjust condition did not deception detection performance. The decision-making strategies of humans suggest that the truth bias - the tendency to assume the other is telling the truth - could explain the detrimental effect. Conclusion: The current study does not support the notion that humans can meaningfully add to the deception detection performance of a machine learning system.
This paper critically examines a recently developed proposal for a border control system called iBorderCtrl, designed to detect deception based on facial recognition technology and the measurement of micro-expressions, termed 'biomarkers of deceit'. Funded under the European Commission's Horizon 2020 programme, we situate our analysis in the wider political economy of 'emotional AI' and the history of deception detection technologies. We then move on to interrogate the design of iBorderCtrl using publicly available documents and assess the assumptions and scientific validation underpinning the project design. Finally, drawing on a Bayesian analysis we outline statistical fallacies in the foundational premise of mass screening and argue that it is very unlikely that the model that iBorderCtrl provides for deception detection would work in practice. By interrogating actual systems in this way, we argue that we can begin to question the very premise of the development of data-driven systems, and emotional AI and deception detection in particular, pushing back on the assumption that these systems are fulfilling the tasks they claim to be attending to and instead ask what function such projects carry out in the creation of subjects and management of populations. This function is not merely technical but, rather, we argue, distinctly political and forms part of a mode of governance increasingly shaping life opportunities and fundamental rights.
We present work on deception detection, where, given a spoken claim, we aim to predict its factuality. While previous work in the speech community has relied on recordings from staged setups where people were asked to tell the truth or to lie and their statements were recorded, here we use real-world political debates. Thanks to the efforts of fact-checking organizations, it is possible to obtain annotations for statements in the context of a political discourse as true, half-true, or false. Starting with such data from the CLEF-2018 CheckThat! Lab, which was limited to text, we performed alignment to the corresponding videos, thus producing a multimodal dataset. We further developed a multimodal deep-learning architecture for the task of deception detection, which yielded sizable improvements over the state of the art for the CLEF-2018 Lab task 2. Our experiments show that the use of the acoustic signal consistently helped to improve the performance compared to using textual and metadata features only, based on several different evaluation measures. We release the new dataset to the research community, hoping to help advance the overall field of multimodal deception detection.
We report our models for detecting age, language variety, and gender from social media data in the context of the Arabic author profiling and deception detection shared task (APDA). We build simple models based on pre-trained bidirectional encoders from transformers (BERT). We first fine-tune the pre-trained BERT model on each of the three datasets with shared task released data. Then we augment shared task data with in-house data for gender and dialect, showing the utility of augmenting training data. Our best models on the shared task test data are acquired with a majority voting of various BERT models trained under different data conditions. We acquire 54.72% accuracy for age, 93.75% for dialect, 81.67% for gender, and 40.97% joint accuracy across the three tasks.
Despite a revolution in the pervasiveness of video cameras in our daily lives, one of the most meaningful forms of nonverbal affective communication, interpersonal eye gaze, i.e. eye gaze relative to a conversation partner, is not available from common video. We introduce the Interpersonal-Calibrating Eye-gaze Encoder (ICE), which automatically extracts interpersonal gaze from video recordings without specialized hardware and without prior knowledge of participant locations. Leveraging the intuition that individuals spend a large portion of a conversation looking at each other enables the ICE dynamic clustering algorithm to extract interpersonal gaze. We validate ICE in both video chat using an objective metric with an infrared gaze tracker (F1=0.846, N=8), as well as in face-to-face communication with expert-rated evaluations of eye contact (r= 0.37, N=170). We then use ICE to analyze behavior in two different, yet important affective communication domains: interrogation-based deception detection, and communication skill assessment in speed dating. We find that honest witnesses break interpersonal gaze contact and look down more often than deceptive witnesses when answering questions (p=0.004, d=0.79). In predicting expert communication skill ratings in speed dating videos, we demonstrate that interpersonal gaze alone has more predictive power than facial expressions.
Most work on automated deception detection (ADD) in video has two restrictions: (i) it focuses on a video of one person, and (ii) it focuses on a single act of deception in a one or two minute video. In this paper, we propose a new ADD framework which captures long term deception in a group setting. We study deception in the well-known Resistance game (like Mafia and Werewolf) which consists of 5-8 players of whom 2-3 are spies. Spies are deceptive throughout the game (typically 30-65 minutes) to keep their identity hidden. We develop an ensemble predictive model to identify spies in Resistance videos. We show that features from low-level and high-level video analysis are insufficient, but when combined with a new class of features that we call LiarRank, produce the best results. We achieve AUCs of over 0.70 in a fully automated setting. Our demo can be found at http://home.cs.dartmouth.edu/~mbolonkin/scan/demo/
We propose a data-driven method for automatic deception detection in real-life trial data using visual and verbal cues. Using OpenFace with facial action unit recognition, we analyze the movement of facial features of the witness when posed with questions and the acoustic patterns using OpenSmile. We then perform a lexical analysis on the spoken words, emphasizing the use of pauses and utterance breaks, feeding that to a Support Vector Machine to test deceit or truth prediction. We then try out a method to incorporate utterance-based fusion of visual and lexical analysis, using string based matching.
Effective and real-time eyeblink detection is of wide-range applications, such as deception detection, drive fatigue detection, face anti-spoofing, etc. Although numerous of efforts have already been paid, most of them focus on addressing the eyeblink detection problem under the constrained indoor conditions with the relative consistent subject and environment setup. Nevertheless, towards the practical applications eyeblink detection in the wild is more required, and of greater challenges. However, to our knowledge this has not been well studied before. In this paper, we shed the light to this research topic. A labelled eyeblink in the wild dataset (i.e., HUST-LEBW) of 673 eyeblink video samples (i.e., 381 positives, and 292 negatives) is first established by us. These samples are captured from the unconstrained movies, with the dramatic variation on human attribute, human pose, illumination condition, imaging configuration, etc. Then, we formulate eyeblink detection task as a spatial-temporal pattern recognition problem. After locating and tracking human eye using SeetaFace engine and KCF tracker respectively, a modified LSTM model able to capture the multi-scale temporal information is proposed to execute eyeblink verification. A feature extraction approach that reveals appearance and motion characteristics simultaneously is also proposed. The experiments on HUST-LEBW reveal the superiority and efficiency of our approach. It also verifies that, the existing eyeblink detection methods cannot achieve satisfactory performance in the wild.
Lies and deception are common phenomena in society, both in our private and professional lives. However, humans are notoriously bad at accurate deception detection. Based on the literature, human accuracy of distinguishing between lies and truthful statements is 54% on average, in other words it is slightly better than a random guess. While people do not much care about this issue, in high-stakes situations such as interrogations for series crimes and for evaluating the testimonies in court cases, accurate deception detection methods are highly desirable. To achieve a reliable, covert, and non-invasive deception detection, we propose a novel method that jointly extracts reliable low- and high-level facial features namely, 3D facial geometry, skin reflectance, expression, head pose, and scene illumination in a video sequence. Then these features are modeled using a Recurrent Neural Network to learn temporal characteristics of deceptive and honest behavior. We evaluate the proposed method on the Real-Life Trial (RLT) dataset that contains high-stake deceptive and honest videos recorded in courtrooms. Our results show that the proposed method (with an accuracy of 72.8%) improves the state of the art as well as outperforming the use of manually coded facial attributes 67.6%) in deception detection.
Automated deception detection (ADD) from real-life videos is a challenging task. It specifically needs to address two problems: (1) Both face and body contain useful cues regarding whether a subject is deceptive. How to effectively fuse the two is thus key to the effectiveness of an ADD model. (2) Real-life deceptive samples are hard to collect; learning with limited training data thus challenges most deep learning based ADD models. In this work, both problems are addressed. Specifically, for face-body multimodal learning, a novel face-focused cross-stream network (FFCSN) is proposed. It differs significantly from the popular two-stream networks in that: (a) face detection is added into the spatial stream to capture the facial expressions explicitly, and (b) correlation learning is performed across the spatial and temporal streams for joint deep feature learning across both face and body. To address the training data scarcity problem, our FFCSN model is trained with both meta learning and adversarial learning. Extensive experiments show that our FFCSN model achieves state-of-the-art results. Further, the proposed FFCSN model as well as its robust training strategy are shown to be generally applicable to other human-centric video analysis tasks such as emotion recognition from user-generated videos.
Humans are the final decision makers in critical tasks that involve ethical and legal concerns, ranging from recidivism prediction, to medical diagnosis, to fighting against fake news. Although machine learning models can sometimes achieve impressive performance in these tasks, these tasks are not amenable to full automation. To realize the potential of machine learning for improving human decisions, it is important to understand how assistance from machine learning models affects human performance and human agency. In this paper, we use deception detection as a testbed and investigate how we can harness explanations and predictions of machine learning models to improve human performance while retaining human agency. We propose a spectrum between full human agency and full automation, and develop varying levels of machine assistance along the spectrum that gradually increase the influence of machine predictions. We find that without showing predicted labels, explanations alone slightly improve human performance in the end task. In comparison, human performance is greatly improved by showing predicted labels (>20% relative improvement) and can be further improved by explicitly suggesting strong machine performance. Interestingly, when predicted labels are shown, explanations of machine predictions induce a similar level of accuracy as an explicit statement of strong machine performance. Our results demonstrate a tradeoff between human performance and human agency and show that explanations of machine predictions can moderate this tradeoff.
Detecting deception in natural language has a wide variety of applications, but because of its hidden nature there are currently no public, large-scale sources of labeled deceptive text. This work introduces the Mafiascum dataset [1], a collection of over 700 games of Mafia, in which players are randomly assigned either deceptive or non-deceptive roles and then interact via forum postings. Over 9000 documents were compiled from the dataset, which each contained all messages written by a single player in a single game. This corpus was used to construct a set of hand-picked linguistic features based on prior deception research, as well as a set of average word vectors enriched with subword information. A logistic regression classifier fit on a combination of these feature sets achieved an average precision of 0.39 (chance = 0.26) and an AUROC of 0.68 on 5000+ word documents. On 50+ word documents, an average precision of 0.29 (chance = 0.23) and an AUROC of 0.59 was achieved. [1] https://bitbucket.org/bopjesvla/thesis/src
Language use reveals information about who we are and how we feel1-3. One of the pioneers in text analysis, Walter Weintraub, manually counted which types of words people used in medical interviews and showed that the frequency of first-person singular pronouns (i.e., I, me, my) was a reliable indicator of depression, with depressed people using I more often than people who are not depressed4. Several studies have demonstrated that language use also differs between truthful and deceptive statements5-7, but not all differences are consistent across people and contexts, making prediction difficult8. Here we show how well linguistic deception detection performs at the individual level by developing a model tailored to a single individual: the current US president. Using tweets fact-checked by an independent third party (Washington Post), we found substantial linguistic differences between factually correct and incorrect tweets and developed a quantitative model based on these differences. Next, we predicted whether out-of-sample tweets were either factually correct or incorrect and achieved a 73% overall accuracy. Our results demonstrate the power of linguistic analysis in real-world deception research when applied at the individual level and provide evidence that factually incorrect tweets are not random mistakes of the sender.