No description available.

With the continuing spread of misinformation and disinformation online, it is of increasing importance to develop combating mechanisms at scale in the form of automated systems that can find checkworthy information, detect fallacious argumentation of online content, retrieve relevant evidence from authoritative sources and analyze the veracity of claims given the retrieved evidence. The robustness and applicability of these systems depend on the availability of annotated resources to train machine learning models in a supervised fashion, as well as machine learning models that capture patterns beyond domain-specific lexical clues or genre-specific stylistic insights. In this thesis, we investigate the role of models for argument structure and argument quality in improving tasks relevant to fact-checking and furthering our understanding of misinformation and disinformation. We contribute to argumentation mining, misinformation detection, and fact-checking by releasing multiple annotated datasets, developing unified models across datasets and task formulations, and analyzing the vulnerabilities of such models in adversarial settings. We start by studying the argument structure's role in two downstream tasks related to fact-checking. As it is essential to differentiate factual knowledge from opinionated text, we develop a model for detecting the type of news articles (factual or opinionated) using highly transferable argumentation-based features. We also show the potential of argumentation features to predict the checkworthiness of information in news articles and provide the first multi-layer annotated corpus for argumentation and fact-checking. We then study qualitative aspects of arguments through models for fallacy recognition. To understand the reasoning behind checkworthiness and the relation of argumentative fallacies to fake content, we develop an annotation scheme of fallacies in fact-checked content and investigate avenues for automating the detection of such fallacies considering single- and multi-dataset training. Using instruction-based prompting, we introduce a unified model for recognizing twenty-eight fallacies across five fallacy datasets. We also use this model to explain the checkworthiness of statements in two domains. Next, we show our models for end-to-end fact-checking of statements that include finding the relevant evidence document and sentence from a collection of documents and then predicting the veracity of the given statements using the retrieved evidence. We also analyze the robustness of end-to-end fact extraction and verification by generating adversarial statements and addressing areas for improvements for models under adversarial attacks. Finally, we show that evidence-based verification is essential for fine-grained claim verification by modeling the human-provided justifications with the gold veracity labels.

No description available.

The effects of simple physical countermeasures on the validity of the control question test (CQT). a physiological detection of deception (PDD) technique. was investigated in a laboratory mock criae paradigm. Forty-eight Introductory Psychology students served as subjects and were assigned to either an innocent group or to one of three guilty groups. who participated in a mock crime. Two of the guilty groups were trained in a countermeasure technique, either self-induced pain or muscle tension, and were coached as to when to produce responses in order to beat the test. All subjects were motivated to produce truthful responses by the offer of credit points toward their final grades if they were classified as truthful on a subsequent PDD examination. All subjects were given a field type CQT examination by an experienced field PDD examiner. The examiner correctly classified 52% of the subjects, incorrectly classified 61% and called 42% inconclusive. Countermeasure usage did not produce changes in the frequency of either inconclusive of incorrect classifications. Statistically significant differences were found between the innocent and guilty groups in the semi-objective scores and in all of the objectively reduced measures (except heart rate). A significant but small difference was found between the guilty control and the countermeasures groups in finger pulse amplitude but in no other objective measure nor in the semi-objective scores. These results suggest the validity of the CQT to be robust in the face of the countermeasures used. Implications for field use of the CQT were discussed. / M. S.

To assess fraud risk, auditors collect evidence in a sequential manner by reviewing workpaper documentation, and by collecting corroborating and clarifying information from financial (management) personnel and nonfinancial (operating) personnel. SAS 99 (AICPA, 2002) noted that audit evidence gathered from financial personnel may be susceptible to deception. In addition, prior researchers have found auditors to be poor at detecting deception immediately following deceptive communication. Though the audit process is sequential and iterative, these studies measured auditors– ability to detect deception at a single point and did not provide corroborating evidence after the deceptive communication for auditors to revise their judgments. In this study, I examined auditors’ fraud risk assessments and truthfulness judgments throughout the audit process when there was an attempt at deception by management (financial) personnel. The belief adjustment model provided a framework to examine auditors’ initial judgments, their judgments directly following a deception attempt by financial personnel, and their judgments after receiving corroborating evidence from nonfinancial personnel. Sixty-four experienced auditors electronically completed one of four randomly assigned cases and, within each case, assessed the fraud risk, truthfulness, trustworthiness, and reliability of financial personnel at multiple points for a fictitious client. I manipulated the presence (absence) of fraud and the level of experience of the source of corroborating evidence (operating personnel). I hypothesized that auditors would not be able to differentially evaluate fraud risk and truthfulness judgments of financial personnel between the fraud and no fraud conditions when exposed to workpaper documentation and deceptive client inquiry evidence by management (financial personnel). However, I expected to find that auditors– would update their fraud risk and truthfulness judgments as they reviewed audit evidence from nonfinancial (operating) personnel. The results indicate that auditors in this study are not able to appropriately assess fraud risk and the truthfulness of financial personnel following the review of workpaper and client inquiry evidence. While the client was deceptive in the fraud condition only, auditors did not differentially assess the fraud risk and truthfulness of financial personnel between the fraud and no fraud conditions. After auditors reviewed evidence from nonfinancial personnel, in the presence of fraud, auditors increased their fraud risk and decreased their truthfulness judgments of financial personnel as inconsistent evidence was presented from a corroborating source. Therefore, in the presence of fraud, auditors improved the effectiveness of the audit process by appropriately increasing their fraud risk assessments in light of inconsistent audit evidence from nonfinancial (operating) personnel. Of equal importance, in the absence of fraud, auditors decreased their fraud risk assessments as consistent evidence was presented from a corroborating source. Therefore, auditors increased the efficiency of the audit process by appropriately decreasing their fraud risk assessments after integrating consistent audit evidence from nonfinancial personnel into their judgments. Further, I observed that these auditors revised their fraud risk assessments to a greater extent when audit evidence was provided by a source with a higher level of experience. Though prior research has found auditors to be poor at detecting deception, the results of this study indicate that auditors will increase or decrease their fraud risk assessments and truthfulness judgments based on the consistency of audit evidence gathered from a corroborating source. Therefore, in practice, auditors may be able to detect deception as the audit progresses. / Ph. D.

A procedure similar to correspondence training was used to assess the propensity of children to lie under varying levels of adult monitoring. Thirteen children selected pieces of food to be eaten later as a snack, and reported their selections to an experimenter. The manipulation involved reinforcement for reports of having selected a previously unselected food. Eleven of 13 subjects were completely honest with no differences between subjects with a history of frequent lying and subjects with a generally honest history. These data are inconsistent with previously published correspondence training studies. Subjects may have responded to environmental cues that served as discriminative stimuli indicating that subjects' behavior was being monitored. Subjects may have entered the study with a generalized correspondence rule which left their behavior insensitive to contingencies in the protocol. Parental reports of frequent motives for lying and behavior problem scores are also reported. Suggested directions for future research within this paradigm include comparisons across levels of cognitive development, incorporating naturalistic settings with more familiar adults, examining influence of live or symbolic modeling of lying, and using more clearly aversive target behaviors. / Master of Science

Children's response to a resistance-to-temptation (RTT) task was investigated under three punishment threat conditions: negative consequence, removing an anticipated reward, and no explicit punishment. Ninety first and second graders participated in the RTT task and seventy-three parents completed the Behavior Assessment System for Children and the Psychopathy Screening Device. As only 4% of children transgressed, results are unclear. Hypotheses tested using approximations of transgression showed no differences in RTT. Children with temperaments characterized by hyperactivity, impulsivity, attention problems, and conduct problems (HIA-CP) had the highest levels of psychopathic traits compared to all others. In addition, spanked children were rated as having significantly more behavioral problems than non-spanked children. Limitations of the current study and suggestions for future research are discussed.

Most of the research conducted to measure credibility has focused on comparing one type of media with another, i.e., newspaper vs. television. Other research has looked into how different target audiences of corporate or company newspapers perceive the credibility of the publication. To date, there has been little research into how the target audiences perceive the credibility of a military installation's command newspaper. This study examined how active-duty Army personnel perceived the credibility of an Army installation's command newspaper; specifically, The Paraglide from Fort Bragg, North Carolina.The study employed the model developed by Meyer (1988) to measure perceptions of credibility. The study also used a model developed by Surlin and Walker (1975) to measure the respondent's self-agreement with how three hypothetical "bad news" stories should and would be covered by the command newspaper. The independent variables for the study were: civilian education level, years of service in the military, and job level/rank. The data was collected using a survey questionnaire distributed to Army units assigned to the 82nd Infantry Division (Airborne). Respondents were directed by their supervisors to complete the survey.The results indicated education level was not significant in perceptions of credibility; years of service was significant in perceptions of credibility; and job level/rank was significant in perceptions of credibility. Additionally, education level, years of service, and job level/rank were all significant in the respondent's self-agreement with how controversial issues were covered, but the significance appeared to be issue dependent. / Department of Journalism

Univ., Diss.--Paris., 2002.

Thesis (Ph. D.)--University of Wisconsin--Madison, 1967. / Vita. eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references.