Bilkent University Institutional Repository (BUIR)
Bilkent University Institutional Repository (BUIR), a service of Bilkent University Libraries, collects, preserves, and distributes the intellectual output of Bilkent University. Faculty, staff, and students are invited to deposit their research and scholarship. Departments, administrative units, programs, and centers are invited to use the Institutional Repository to distribute their working papers, technical reports, conference proceedings, and other research material. For assistance in depositing documents. For more information, please contact us.
Scholarly Publications
University Library
Bilkent Theses
Recent Submissions
Spatial experience in metaverse exhibitions: a comparative analysis of reality-based and virtuality-based design approaches
(Bilkent University, 2024-07) Uzun, Zeynep
Technological advancements have initiated the realisation of the metaverse concept, which provides an alternative environment for human experience. Within this digital frontier, while some architectural designs transcend the boundaries of reality ` adaptation, certain spaces adhere to mimicking the physical realm. As a consequence, two design approaches have emerged in the metaverse: reality-based (RB), which replicates the appearance of real-life architecture, and virtuality-based (VB), which takes advantage of the flexibility and limitlessness of virtual environments. However, the impact of these different approaches to the design on spatial experience remain to be investigated.
This thesis aims to find out how these two different approaches in metaverse environment’s designs impact spatial experience of the users. By concentrating on two different exhibition spaces in metaverses, the effect of architectural design approaches on spatial experience is intended to be find out. Within this framework, an experimental study was conducted with 118 architectural design students. The spatial experience was analysed through the Questionnaire on Spatial experience in exhibition environments with 6 subcategories (presence, immersion, engagement, flow, judgement, and wayfinding) through a 5-point Likert Scale. As a result of the research, the participants who experienced the RB environment felt significantly more present than those who experienced the VB environment. The result aid in comprehending the impact of design on spatial experience by providing insights from the perspectives of architecture, art, and spatial experience. Future research may validate and expand upon these discoveries in a more comprehensive framework through the utilisation of spaces featuring diverse design methodologies, and in collaboration with individuals from other disciplines.
Investigation of the neural basis of natural action perception within the theoretical framework of perceptual decision making
(Bilkent University, 2024-07) Evsen, Şeyda
Every day, we perceive many actions performed by people around us. Understanding the movements we perceive and the intentions behind these movements is critical to determining how to respond to the observed action. Neurophysiology studies and computational model theories suggest that this perceptual decision process consists of two stages: accumulation of encoded sensory evidence and selection of response. EEG studies in humans have identified a neural trace thought to represent the first of these two stages: Centro Parietal Positivity (CPP), which scales with the strength of the incoming sensory evidence and differs from the Lateralized Readiness Potential (LRP) that represents motor preparatory activity. However, these studies mainly utilized simple and artificial stimuli; therefore, whether perceptual decision-making processes will be similar when processing a more complex and ecologically meaningful stimulus, such as natural actions, is an open question. This study aimed to investigate this gap in the literature by examining the neural basis of perceptual decisions related to the perception of natural actions. To this end, in this study, which included three separate EEG sessions, twenty participants performed a discrimination task between action exemplars belonging to a different action class in each session (either one of the locomotion, skin-displacement actions, or manipulation actions). To control for task difficulty, the same coherence manipulation was applied to each action exemplar to create four levels of coherence, and behavioral data analysis showed that mean response times and miss rates decreased as the level of coherence increased. Furthermore, the results of event-related potential analysis showed that the CPP signal followed the level of coherence of the stimulus within all action classes, which is in line with the literature, and also that the manipulation of the action class had a significant effect on the CPP signal. In contrast to CPP, the LRP signal showed an independent build-up from both the strength of the sensory evidence and the identity of the action class. Taken together, the findings of the current study support the generalizability of perceptual decision-making stages and their neural basis, which are defined by utilizing much simpler stimuli, to decisions related to the perception of natural actions while revealing that the identity of natural movements significantly influences the decision process.
Effective connectivity in cortical regions during bottom-up perception of biological motion under attentional load: an FMRI-DCM study
(Bilkent University, 2024-07) Mert, Sezan
The ability to detect biological motion holds an evolutionarily important role in vital and social functions. However, in our daily lives, we perceive biological motion while we are at a task most of the time. In other words, it is perceived when our attention is directed at another thing. In this aspect, understanding the dynamics of its bottom-up perception is of high importance. Meanwhile, the attentional mechanisms and where their effects occur are a matter of debate in the literature, sparking off various theories, such as early selection, late selection, and attentional load theory. Dynamic causal modeling (DCM) is a suitable tool for investigating the dynamics of attentional effects on the network, enabling the bottom-up perception of biological motion, and comparing the existing theories in the literature with Bayesian graph models. To this end, we utilized the DCM approach with fMRI data collected using an attentional load paradigm and biological motion peripheral distractors [1]. In our model space, we modeled the theories of selective attention along with two complementary models. The Bayesian Model Selection (BMS) showed that the model that explained the data the best was the model where both attentional load conditions modulated all top-down connections rather than the models of existing theories. This showed that attentional effects take part in the bottom-up perception, not in a focused location, such as early or late, but in a more distributed manner throughout the processing pipeline. Further statistical tests on the model parameters yielded no difference between load conditions and between biological motion and scrambled motion in their modulation strengths. Yet, the strengths of biological motion on different connections were different from each other. A similar observation is also made for the low load condition but not for the scrambled motion and high load conditions. The former can be accepted as evidence for the differential processing of biological and scrambled motion. The latter may be explained by a spillover of perceptual resources on biological motion and causing competition in low-load conditions.
The effects of smoking restrictions on class attendance and life satisfaction: evidence from a Turkish University
(Bilkent University, 2024-06) Kılınç, Sena
We investigate the effect of smoking regulations on a university campus regarding attendance frequency, life satisfaction, and campus satisfaction among university students. A private university in Turkey implemented a campus-wide smoke-free policy, which prohibited smoking on campus except in designated areas at least 150 meters from faculty buildings. In our theoretical framework, we analyze a simple game for smoker and nonsmoker students with different types. In order to analyze the testable implications, we conducted an online survey with retrospective questions to third and fourth-year students. 406 students responded to our survey. In the empirical methodology, we model attendance frequency as a function of various factors, including individual fixed effects driven by intrinsic motivation for attendance, faculty fixed effects affected by faculty-related attendance policies, a time effect that may impact attendance trends, and a heterogeneous policy effect based on time and distance to smoking areas. As an alternative approach to the first one, instead of modelling individual fixed effects and adjusting for them through first differencing, we employ a difference-in-differences model. Our findings indicate that attendance increased and the smoking intensity among students decreased after the implementation of the policy for smoker students. Remarkably, the results obtained through the alternative approach align with those of the primary methodology. Also, the findings are consistent with the theoretical framework. JEL Classification: I1, I23, I28, I31.
The effect of fluid viscoelasticity in soft elastohydrodynamic lubrication
(Bilkent University, 2024-07) Sarı, Mehmet Hakan
Lubrication is crucial across various industries and natural environments to prevent direct contact between solid surfaces. Elastohydrodynamic lubrication specifically addresses situations where solid surfaces are likely to deform. Viscoelastic additives in fluids can alter lubrication characteristics significantly. Reduced models available today efficiently solve viscoelastic equations numerically in thin film contacts, surpassing direct numerical computations in efficiency. Employing numerical techniques, viscoelastic fluids within the linear elastic regime are successfully analyzed. Further investigations into contraction geometries aim to elucidate findings on soft lubrication. An elementary experimental study was conducted to demonstrate the effect of viscoelasticity on load-carrying capacity. A comprehensive literature review provided the necessary mathematical foundation for understanding viscoelasticity and solid deformation. Concepts such as the upper convected time derivative and polymeric constitutive equations were summarized. Reynolds equation, incorporating polymeric elastic stress, is combined under the thin film lubrication approximation. Viscoelasticity is characterized by the non-dimensional number De, which represents the ratio of polymer relaxation time to observation time t/to . In the context of lubrication, the observation time t/to is defined as L/U, where L is the length of the channel and U is the speed of flow. The reduced numerical methods are based on finite differencing schemes and linearization of the flow with respect to De. The linearized flow using the Oldroyd-B constitutive model forms the basis for a more complex numerical model that is compatible with direct numerical solvers. A unique semi-implicit method has been developed to solve nonlinear stress equations. The coupling between solid and fluid solvers is fully explained with schematics, and the boundary element method is integrated with the finite difference method. Prior to presenting results, the developed numerical methods were validated against previous outcomes published in journal articles. Elastohydrodynamic lubrication results indicate that the friction coefficient decreases with increasing De at high deformabilities. To further illustrate EHL cases, step-like channels were studied. Finally, experimental data is provided for future use, concluding with remarks to contextualize our findings.
Prescriptive modeling for counterfactual inferences
(Bilkent University, 2024-06) Işık, Elif Sena
In real-life scenarios, conducting experiments or simulations to optimize out-comes can be costly in terms of time and resources. This thesis explores the utilization of trained neural networks for predictive modeling and optimization to address this challenge. The methodology involves training neural networks on historical data or simulated environments to capture complex relationships be-tween input variables and outputs. We then employ optimization techniques to explore parameter/input spaces and identify optimal configurations for desired outputs. Importantly, this approach enables us to conduct counterfactual analyses, allowing us to assess how changes in input parameters would affect outputs. We present case studies utilizing two distinct real-life scenarios: firstly, the public simulation model FluTE, where we demonstrate the effectiveness of our approach in optimizing strategies to alleviate the spread of infectious diseases. Secondly, we tackle an assortment problem and demonstrate how decision-making processes in retail settings can be assisted by trained neural networks to maximize profitability. We then also suggest an improved methodology to control the uncertainty in predicted outputs from neural network. We utilize dropout networks to quantify variability in the output predictions and embed them into the optimization model. Computational experiments are conducted with the two case studies and customized problem specific methodologies are suggested that includes decomposition methods and heuristics.
Robust optimization models for network revenue management
(Bilkent University, 2024-07) Bahtiyar, Irem
Effective capacity allocation methods play a crucial role in Network Revenue Management. Yet, current methods for determining optimal capacity controls under uncertainty, such as stochastic optimization, often assume a known probability distribution for unknown parameters. This assumption may degrade a model’s performance when faced with unexpected data patterns. This thesis explores a novel approach through robust optimization to address stochastic resource allocation problems. We introduce a heuristic based on these robust formulations to derive actionable results. Through extensive simulations focused on seat allocation problems within the revenue management domain, our proposed formulations demonstrate improved worst-case performances. Notably, even under favorable scenarios, our solutions remain comparable to existing methods in the revenue management literature.
Power as visibility: conceptualizing the changing nature of influence in international relations
(Bilkent University, 2024-06) Karabıçak, Onur
This thesis propounds a new definition of power and influence in international relations and a conceptual framework of power’s translation into influence. First, this thesis defines power as being or rendering something (in)visible, and influence as changing one's narrative of world politics. Visibility is being socially relevant and mattering to others, and it is operationalized as an actor's frequency of being seen by an audience. Second, this thesis draws a framework for the analysis of the power and influence of international actors on audiences. The methodology is extracted from the operationalization of the components of visibility in the proposed framework of analysis. The power as visibility framework is applied to the case of comparing Erdoğan's power and influence over the mass audience in Türkiye and the Western European elite audience in 2008 and 2020. The material examined by the thesis consists of official speeches and popular culture media. Finally, the thesis concludes with a discussion of the alternative explanations and power as visibility framework.
Multifaceted analysis of older adults’ and caretakers’ attitudes toward social robots
(Bilkent University, 2024-07) Çonka, Begüm
The global aging population is increasing drastically, bringing crucial concerns regarding the lives of older adults. Though older people generally prefer staying at their own homes while aging, it’s necessary to find solutions that will enhance their well-being and life quality since they experience many psychological and physical problems. This thesis focused on social robots’ role in enriching older adults’ environments, thus improving their healthspan by enhancing their psychological and physiological health. Therefore, aim of this thesis was to understand older adults’ and caretakers’ attitudes toward social robots. Semi-structured interviews were conducted with older adults (N=18) and caretakers (N=12). In the interviews, photos and videos of the three social robots with different abilities, appearances, and human-likeness levels were shown. These robots are the pet- like robot Aibo, the toy-like robot Paro, and the humanoid robot Pepper. Then, participants were asked to answer open-ended questions to explore their feelings about the robots, their preferences regarding meeting and interacting with them, and whether they would want to have these robots in their homes. Results were analyzed using sentiment analysis, which is a Natural Language Processing (NLP) method, and a qualitative analysis method, thematic analysis (TA). Sentiment analysis results demonstrate a differentiation in caretakers’ attitudes toward social robots. They perceived Paro negatively but responded positively to Aibo and Pepper. On the other hand, older adults seem to have similar attitudes toward the three robots. Three experts working in the field of human-robot interaction conducted the TA separately. Combining their analysis, a new model was created, and when all experts confirmed this model, four overarching themes emerged: I) Perceived and expected roles of the robot, II) Physical characteristics and de- sign features of the robot, III) Factors influencing acceptance of the robot, and IV) Disadvantages of the robot. Themes and sub-themes under the overarching themes differed partly between older adults and caretakers and across Aibo, Paro, and Pepper. The findings of this thesis contribute to the literature by comparing the attitudes toward three different social robots in Turkish culture with the adoption of a multifaceted design approach.
Don’t be average: distinguishing benign social comparisons from mastery goals and harsh social comparisons
(Bilkent University, 2024-06) Akay, Enise
This thesis explores the concept of benign social comparisons within the framework of achievement goal theory (Dweck, 1986) and their impact on motivational factors including perceived competence, psychological pressure, usefulness, and enjoyment, alongside behavioral measures such as effort and performance. Benign social comparisons, an unexplored area, propose an alternative approach where individuals aim to perform at an average level rather than striving to surpass (any) other, as traditionally implied through performance-approach goals. The study consists of two experiments involving computerized tangram puzzles. The first study investigated whether benign social comparison differed from harsh social comparisons (traditional performance goal approach) and mastery goals in relation to motivational and behavioral outcomes. The results revealed that participants reported higher perceived competence and usefulness in the mastery goal condition compared to the benign social comparison condition. Additionally, participants in the harsh social comparison condition reported greater usefulness compared to those in the benign social comparison condition. The second study examined whether feedback would moderate the effects of social comparisons versus mastery goals. Although the results revealed no main effects of feedback, an interaction effect between conditions and feedback was found regarding usefulness. Participants reported greater usefulness in the mastery goal condition than in the benign social comparison condition when positive feedback was provided. Additionally, participants reported experiencing less psychological pressure, perceiving greater usefulness, and investing more effort in mastery goals compared to benign social comparisons. These findings underscore the non-adaptive nature of benign social comparisons compared to mastery goals, contradicting initial hypotheses. Surprisingly, benign comparisons may be even more detrimental than harsh social comparisons. However, these findings underscore the need for additional research in this field.