Usage of language models in an in-context learning environment has been adapted for a wide range of tasks. Recent works have showcased the impact of pretraining data on the in-context performance of language models. In this work, we experiment with numbers having high and low frequencies in the pretraining data to understand the impact of term frequencies on the model's performance. We also experiment with random and adversarial demonstrations to understand the pretraining bias present in the model. Through these experiments, we showcase the importance of pretraining frequencies of the numbers present in the demonstrations and explain how highly frequent terms can be used in the demonstrations to achieve better task performance. Moreover, we also show the impact of pretraining bias on the model's performance and explain how the model overcomes this bias with more demonstrations. / Master of Science / Recent works focus on understanding and improving the arithmetic capabilities of the state-of-the-art (SOTA) systems in the domain of Natural Language Processing (NLP). This work focuses on designing and performing novel experiments to analyze the impact of training data on the performance of such systems. Through these experiments, this work showcases interesting properties of the SOTA systems which will promote future research to understand them better as well as help in creating better downstream applications.

Identifying whether a stimulus is threatening or not is critical for staying safe. The faster one can detect a threat, the greater chance there is to avoid any potential danger. Factors contributing to the visual attention of threat are therefore informative. Previous research has examined how aspects of temperament and effortful control interact and affect the attention allocated to threats, especially in clinically anxious populations. However, there is a sparsity of this literature existing for nonclinical populations. My study addressed previous gaps by examining whether negative affect and fear impact an attention bias to threat in children aged 6 through 8 while assessing how attentional control and inhibitory control moderate these relations. A modified visual search task with snakes as the threat was given to the participants after the children’s parents completed questionnaires and the children completed an attentional control task. Results showed that an attentional bias to snakes was seen in the sample. Negative affect as a main effect nor as an interaction effect with attentional control predicted for the attention bias to snakes. Fear predicted for the attention bias to snakes as a main effect. Interestingly, inhibitory control moderated the relation between fear and the attention bias to snakes. Only children with high inhibitory control and high fear predicted for the attention bias to snakes. Findings may indicate children with this temperament are more vulnerable to the onset of anxiety. / M.S. / Identifying whether a stimulus is threatening or not is critical for staying safe. The faster one can detect a threat, the greater chance there is to avoid any potential danger. Factors contributing to the visual attention of threat are therefore informative. Previous research has examined how aspects of temperament and effortful control interact and affect the attention allocated to threats, especially in clinically anxious populations. However, there is a sparsity of literature existing for nonclinical populations. My study addressed previous gaps by examining whether aspects of temperament, specifically negative affect and fear, impact an attention bias to threat in children aged 6 through 8 while assessing how aspects of effortful control, specifically attentional control and inhibitory control, moderate these relations. A visual search task where participants would select a target among distractors with snakes as the target representing threat was given to the child participants after the children’s parents completed questionnaires and the children completed an I-spy task which measured the children’s attentional control. Results showed that an attentional bias to snakes was seen in the sample. Negative affect did not solely nor when interacted with attentional control predict for the attention bias to snakes. Fear predicted for the attention bias to snakes as a main effect. Interestingly, inhibitory control moderated the relation between fear and the attention bias to snakes, which meant that only children with high inhibitory control and high fear predicted for the attention bias to snakes. Findings may indicate children with this temperament are greater susceptible the development of anxiety.

Yes / The purpose of this study was to isolate and quantify the effects of observer response criterion on perimetric sensitivity, response variability, and maximum response probability. Twelve people with glaucoma were tested at three locations in the visual field (age = 47-77 years, mean deviation = -0.61 to -14.54 dB, test location Humphrey field analyzer [HFA] sensitivities = 1 to 30 dB). Frequency of seeing (FoS) curves were measured using a method of constant stimuli with two response paradigms: a "yes-no" paradigm similar to static automated perimetry and a criterion-free two interval forced choice (2IFC) paradigm. Comparison measures of sensitivity, maximum response probability, and response variability were derived from the fitted FoS curves. Sensitivity differences between the tasks varied widely (range = -11.3 dB to 21.6 dB) and did not correlate with visual field sensitivity nor whether the visual field location was in an area of steep sensitivity gradient within the visual field. Due to the wide variation in differences between the methods, there was no significant difference in mean sensitivity between the 2IFC task relative to the yes-no task, but a trend for higher sensitivity (mean = 1.9 dB, SD = 6.0 dB, P = 0.11). Response variability and maximum response probability did not differ between the tasks (P > 0.99 and 0.95, respectively). Perimetric sensitivity estimates are demonstrably altered by observer response criterion but the effect varies widely and unpredictably, even within a single test. Response bias should be considered a factor in perimetric test variability and when comparing sensitivities to nonperimetric data. The effect of response criterion on perimetric response variability varies widely and unpredictably, even within a single test. / Supported by ARC LP130100055; ARC LP150100815 (AT and AMM), College of Optometrists Research Fellowship (JD).

This thesis explores an approach to predicting recidivism by leveraging crowdsourcing, contrasting traditional judicial discretion and algorithmic models. Instead of relying on judges or algorithms, participants predicted the likelihood of re-offending using the COMPAS dataset, which includes demographic and criminal record information. The study analyzed both quantitative and qualitative data to assess biases in human versus algorithmic predictions. Findings reveal that homogeneous crowds reflect the biases of their composition, leading to more pronounced gender and racial biases. In contrast, heterogeneous crowds, with equal and random distributions, present a more balanced view, though underlying biases still emerge. Both gender and racial biases influence how re-offending risk is perceived, significantly impacting risk evaluations. Specifically, crowds rated African American offenders as less likely to re-offend compared to COMPAS, which assigned them higher risk scores, while Caucasian and Hispanic offenders were perceived as more likely to re-offend by crowds. Gender differences also emerged, with males rated as less likely to re-offend and females as more likely. This study highlights crowdsourcing's potential to mitigate biases and provides insights into balancing consistency and fairness in risk assessments. / Master of Science / Within the criminal justice system, predicting whether someone will re-offend has typically depended on the judgment of judges and computerized systems. This thesis investigates another avenue for predicting re-offending by using crowdsourcing, which gathers input from a group of people. In this study, participants were asked to predict the likelihood of re-offending for several offenders using demographic and criminal record information from the publicly available COMPAS dataset. Participants provided scores, and some also explained their reasoning. Bias, defined as a systematic unfairness that leads to prejudiced outcomes, was a key focus. To understand bias, the study created different groups within the participant crowd based on age, gender, and race, and compared their predictions with COMPAS scores. The analysis revealed important insights into the biases present in both human and algorithmic predictions. A homogeneous crowd, is associated with minimal differences in ratings across genders and races, suggesting a consistent but potentially biased perspective. While a diverse crowd, leads to varied ratings without a clear trend, reflecting a broader range of viewpoints but also increased variability. This suggests that while a diverse crowd may help reduce bias, it can also result in less predictable assessments.

No / Radiographers and other healthcare professionals are becoming increasingly involved in radiological reporting, for instance plain radiographs, mammography and ultrasound. Systematic reviews of research evidence can help to assimilate a knowledge base by ordering and evaluating the available evidence on the reporting accuracy of different professional groups. This article reviews the biases that can undermine the results of plain ¿lm reading performance studies. These biases are subdivided into three categories. The ¿rst category refers to the selection of subjects, including both ¿lms and professionals, and covers the validity of generalizing results beyond the study population. The other two categories are concerned with study design and the interpretation both of ¿lms and of reports and the effect on study validity. An understanding of these biases is essential when designing such studies and when interpreting the results of existing studies.

Is home biasness common among modern investors? To which extent do Swedish investors diversify their investments on an international level? Does home biasness negatively affect the investors performance? To answer these questions, correlation tests of various international indices ranging over four different time periods are conducted, in order to see if correlation between markets are stronger today than before, as stronger correlation would render diversification less useful. To enhance the study, the holdings of the top ten Swedish funds, measured in fund capital according to Morningstar, is reviewed, based on data collected per 2014-12-31 from the Swedish Financial Supervisory Agency (FI). This gives an overview of how the funds diversify their investments internationally, these funds will in turn represent the average Swedish investor in the thesis. By constructing a bullet curve from a set of international indices, the author will analyse to which grade international diversification is useful. The results are that international diversification isn’t as beneficial as theory suggests it is. The reason for it may be due to stronger correlation between international markets in the past 15 years. Most of the Swedish funds tends to be rather home biased in their investments, as about a quarter of the holdings usually are placed in Swedish assets, and in accordance with the results of the indices development, the more home biased they are to Sweden, the better they tend to perform. / Hur vanligt är egentligen home biasness hos den moderna investeraren? Till vilken grad diversifierar egentligen den vanliga Svenska investeraren sina tillgångar internationellt? Påverkar en eventuellt inhemskt investeringsfilosofi investeraren negativt? Korrelationstester för olika världsindex kommer utföras, så att en överskådade blick kan fås över hur världsmarknader rör sig allt mer symmetriskt, då starkare symmetri mellan marknader minskar nyttan av internationell diversifiering. Data om innehav från Sveriges topp tio fonder, sett till fondförmögenhet utifrån Morningstar, har samlats från finansinspektionen per 2014-12-31. Dessa fonder ska representera den typiske Svenska investeraren och dess diversifieringsvanor. Genom att ha samlat data från ett flertal internationella index har effektiv front samt en fiktiv kombination av index skapats för att få fram huruvida avkastning i relation till risk ökar genom internationell diversifiering. Denna kombination av index jämförs sedan mot utveckling av en handfull internationellt samlade index för att se om diversifiering förbättrar avkastningen i relation till risk. Resultatet säger att det index som är mest diversifierade inte är så gynnsamt som teorin påstår. Anledningen till detta kan bero på den ökade korrelationen bland aktiemarknader idag jämfört med för 15 år sedan. De flesta Svenska fonderna har en större andel av sitt innehav i Svenska värdepapper och överlag, förutom diverse undantag, så har det gynnat dessa fonder i avkastning sett till risk.

In 2010 the Texas Education Agency adopted newly rewritten curriculum standards for the state's social studies courses K-12. The period leading up to the adoption of the standards proved contentious as the new standards moved out of the writing committees and into the public realm. The issues brought forth from initial readings of the standards appeared in the national spotlight as educators and the general public from around the country voiced their concerns. Many concerns centered on the belief that the Texas State Board of Education had rewritten the standards to reflect a more politically conservative world view along with an emphasis on traditional Judeo-Christian beliefs. Members of the state board argued that the previous standards reflected a politically liberal world view and that the board needed to bring balance to the standards. Because of national criticism changes were made but on a limited scale. The purpose of this study was to assess if the rewritten standards demonstrated a politically conservative preference as well as an emphasis on Judeo-Christian beliefs. During the analysis I included information regarding a preference for capitalism, a focus on Texas and U.S. history while minimizing world history and a focus on memorization rather than critical thinking. The analysis revealed that the rewritten standards demonstrate a politically conservative leaning while promoting traditional Judeo-Christian beliefs. Analysis and research further revealed that the standards promote capitalism while limiting the study of other economic systems in the world. In regards to the preference for Texas and U.S. history versus world history my analysis confirmed a limited presence of world history; however, research also revealed that this issue is not limited to the state of Texas, but a number of other states in the U.S. Finally, my analysis demonstrated that many of the standards' objectives begin with verbs found on verb lists from Bloom's Taxonomy. By using Bloom's Taxonomy I learned that a majority of objectives throughout the K-12 standards operate in the lower half of the Taxonomy rather than the upper half meaning that there is an emphasis on lower order thinking skills rather than higher order thinking skills.

Die vorliegende Arbeit untersucht Zusammenhänge zwischen verschiedenen Vulnerabilitätsfaktoren der Depression und deren Einflüsse auf den Krankheitsverlauf bei stationär behandelten, depressiven Patienten.

The goals of revision for the Wechsler Intelligence Scale for Children-Third Edition included enhancement of the factor structure, improvement of subtests, and revision of norms. The researchers reported that the very few items that were found to be biased were replaced. The WISC-III performance of a clinical sample of African American, Caucasian, and Hispanic children was analyzed to determine if the test bias was eliminated as claimed in the goals of the revision.

Two procedures for detection of differential item functioning (DIF) for polytomous items were generalized to detection of differential testlet functioning (DTLF). The methods compared were the logistic discriminant function analysis procedure for uniform and non-uniform DTLF (LDFA-U and LDFA-N), and the Mantel score test procedure. Further analysis included comparison of results of DTLF analysis using the Mantel procedure with DIF analysis of individual testlet items using the Mantel-Haenszel (MH) procedure. Over 600 chi-squares were analyzed and compared for rejection of null hypotheses. Samples of 500, 1,000, and 2,000 were drawn by gender subgroups from the NELS:88 data set, which contains demographic and test data from over 25,000 eighth graders. Three types of testlets (totalling 29) from the NELS:88 test were analyzed for DTLF. The first type, the common passage testlet, followed the conventional testlet definition: items grouped together by a common reading passage, figure, or graph. The other two types were based upon common content and common process. as outlined in the NELS test specification.