arXiv subjects

Gizem Gezici

Publications and source records attributed to Gizem Gezici.

At least 19 recordsLinked to original sources

arXiv2026

From Plausible to Actionable: A Position on LLM Self-Explanations

Large Language Models (LLMs) can generate natural language explanations that rationalize their own decisions, a phenomenon commonly referred to as self-explanations. Such explanations have emerged as a promising direction for explainable artificial intelligence (XAI), particularly for interpreting LLM behavior. However, while self-explanations often appear plausible, whether they faithfully reflect a model's underlying reasoning process remains an open question. In this opinion paper, we argue that self-explanations can be highly plausible, questionably faithful, and yet highly actionable. From a traditional XAI perspective, we identify the limitations of standard evaluation protocols for LLM-generated self-explanations and propose practical guidelines for assessing their plausibility and faithfulness. Moreover, we argue that evaluation should extend beyond these criteria to actionability, highlighting applications of LLM rationalization capabilities that support informed decision-making and appropriate action across diverse stakeholders.

cs.CL↗

arXiv2026

Learning by Surprise: Adaptive Mitigation of Model Collapse in Large Language Models

As AI-generated content increasingly populates the web, generative AI models are at growing risk of being trained on their own outputs, a process known as AI autophagy. This feedback loop has been shown to induce model collapse, typically characterized by a loss of diversity in generated content. However, existing work offers a limited understanding of this phenomenon and relies on mitigation strategies that assume access to human-authored data. In this paper, we conduct extensive simulations across multiple datasets and LLMs to address key gaps in the study of model collapse. First, we introduce model-intrinsic measures based on next-token probability distributions, showing that model collapse corresponds to an increasing concentration of probability mass on a small set of tokens. Second, we demonstrate that model collapse is also associated with a loss of common sense, as measured by a decline in commonsense inference accuracy. Third, we identify perplexity (a measure of model "surprise") as a key driver of collapse: fine-tuning on the least "surprising" documents leads to more severe degeneration. Building on this insight, we propose a perplexity-based filtering strategy that prioritizes high-surprise documents during fine-tuning. Unlike existing approaches, our method does not require distinguishing between human-authored and AI-generated content. Across datasets and LLM families, this strategy consistently mitigates model collapse, achieving performance comparable to, and in some cases better than, human-data baselines, while substantially reducing the concentration of next-token probabilities. Overall, our results provide a unified, model-centric understanding of model collapse and suggest practical, scalable strategies for training generative AI systems in increasingly synthetic environments.

cs.CL↗

arXiv2026

Disagreeing Rationales: Rethinking Classification and Explainability Evaluation in Hate Speech Detection

Human disagreement is ubiquitous and well-known in labeling. However, variation in explanations, captured through token-level human rationales, remains far less explored. At the same time, it is unclear how to best evaluate human labels and rationales -- or even how to best aggregate rationales beyond majority vote -- in light of this variation. Yet, rationales may provide additional insights into the richness of human reasoning, that may differ in style, values and interpretations -- especially in subjective NLP tasks like hate speech detection. In this work, we unify diverse models, training strategies, loss functions, and existing evaluation metrics under a single protocol by systematically re-implementing them across different label and rationale representation spaces. Classification metrics are organized around two key properties -- predictive and distributional -- while explainability metrics through three complementary dimensions: plausibility, faithfulness, and complexity. In this unified supervision framework, we evaluate model behavior across classification and explainability metrics, as well as metric sensitivity to the choice of label (hard and soft) and rationale representation space (hard, intermediate and soft). Results show that both hard and soft metrics favor softer representations, highlighting their effectiveness in capturing variation and the need to rethink evaluation in subjective NLP.

cs.CL↗

arXiv2026

Perspectives in Play: A Multi-Perspective Approach for More Inclusive NLP Systems

In the realm of Natural Language Processing (NLP), common approaches for handling human disagreement consist of aggregating annotators' viewpoints to establish a single ground truth. However, prior studies show that disregarding individual opinions can lead can lead to the side effect of underrepresenting minority perspectives, especially in subjective tasks, where annotators may systematically disagree because of their preferences. Recognizing that labels reflect the diverse backgrounds, life experiences, and values of individuals, this study proposes a new multi-perspective approach using soft labels to encourage the development of the next generation of perspective aware models, more inclusive and pluralistic. We conduct an extensive analysis across diverse subjective text classification tasks, including hate speech, irony, abusive language, and stance detection, to highlight the importance of capturing human disagreements, often overlooked by traditional aggregation methods. Results show that the multi-perspective approach not only better approximates human label distributions, as measured by Jensen-Shannon Divergence (JSD), but also achieves superior classification performance (higher F1 scores), outperforming traditional approaches. However, our approach exhibits lower confidence in tasks like irony and stance detection, likely due to the inherent subjectivity present in the texts. Lastly, leveraging Explainable AI (XAI), we explore model uncertainty and uncover meaningful insights into model predictions.

cs.CL↗

arXiv2025

A survey on the impacts of recommender systems on users, items, and human-AI ecosystems

Recommendation systems and assistants (in short, recommenders) influence through online platforms most actions of our daily lives, suggesting items or providing solutions based on users' preferences or requests. This survey systematically reviews, categories, and discusses the impact of recommenders in four human-AI ecosystems -- social media, online retail, urban mapping and generative AI ecosystems. Its scope is to systematise a fast-growing field in which terminologies employed to classify methodologies and outcomes are fragmented and unsystematic. This is a crucial contribution to the literature because terminologies vary substantially across disciplines and ecosystems, hindering comparison and accumulation of knowledge in the field. We follow the customary steps of qualitative systematic review, gathering 154 articles from different disciplines to develop a parsimonious taxonomy of methodologies employed (empirical, simulation, observational, controlled), outcomes observed (concentration, content degradation, discrimination, diversity, echo chamber, filter bubble, homogenisation, polarisation, radicalisation, volume), and their level of analysis (individual, item, and ecosystem). We systematically discuss substantive and methodological commonalities across ecosystems, and highlight potential avenues for future research. The survey is addressed to scholars and practitioners interested in different human-AI ecosystems, policymakers and institutional stakeholders who want to understand better the measurable outcomes of recommenders, and tech companies who wish to obtain a systematic view of the impact of their recommenders.

cs.IR↗

arXiv2025

Multi-Perspective Stance Detection

Subjective NLP tasks usually rely on human annotations provided by multiple annotators, whose judgments may vary due to their diverse backgrounds and life experiences. Traditional methods often aggregate multiple annotations into a single ground truth, disregarding the diversity in perspectives that arises from annotator disagreement. In this preliminary study, we examine the effect of including multiple annotations on model accuracy in classification. Our methodology investigates the performance of perspective-aware classification models in stance detection task and further inspects if annotator disagreement affects the model confidence. The results show that multi-perspective approach yields better classification performance outperforming the baseline which uses the single label. This entails that designing more inclusive perspective-aware AI models is not only an essential first step in implementing responsible and ethical AI, but it can also achieve superior results than using the traditional approaches.

cs.CL↗

arXiv2025

Bridging the Gap: In-Context Learning for Modeling Human Disagreement

Large Language Models (LLMs) have shown strong performance on NLP classification tasks. However, they typically rely on aggregated labels-often via majority voting-which can obscure the human disagreement inherent in subjective annotations. This study examines whether LLMs can capture multiple perspectives and reflect annotator disagreement in subjective tasks such as hate speech and offensive language detection. We use in-context learning (ICL) in zero-shot and few-shot settings, evaluating four open-source LLMs across three label modeling strategies: aggregated hard labels, and disaggregated hard and soft labels. In few-shot prompting, we assess demonstration selection methods based on textual similarity (BM25, PLM-based), annotation disagreement (entropy), a combined ranking, and example ordering strategies (random vs. curriculum-based). Results show that multi-perspective generation is viable in zero-shot settings, while few-shot setups often fail to capture the full spectrum of human judgments. Prompt design and demonstration selection notably affect performance, though example ordering has limited impact. These findings highlight the challenges of modeling subjectivity with LLMs and the importance of building more perspective-aware, socially intelligent models.

cs.CL↗

arXiv2025

Embracing Diversity: A Multi-Perspective Approach with Soft Labels

Prior studies show that adopting the annotation diversity shaped by different backgrounds and life experiences and incorporating them into the model learning, i.e. multi-perspective approach, contribute to the development of more responsible models. Thus, in this paper we propose a new framework for designing and further evaluating perspective-aware models on stance detection task,in which multiple annotators assign stances based on a controversial topic. We also share a new dataset established through obtaining both human and LLM annotations. Results show that the multi-perspective approach yields better classification performance (higher F1-scores), outperforming the traditional approaches that use a single ground-truth, while displaying lower model confidence scores, probably due to the high level of subjectivity of the stance detection task.

cs.CL↗

arXiv2025

Hybrid Retrieval for Hallucination Mitigation in Large Language Models: A Comparative Analysis

Large Language Models (LLMs) excel in language comprehension and generation but are prone to hallucinations, producing factually incorrect or unsupported outputs. Retrieval Augmented Generation (RAG) systems address this issue by grounding LLM responses with external knowledge. This study evaluates the relationship between retriever effectiveness and hallucination reduction in LLMs using three retrieval approaches: sparse retrieval based on BM25 keyword search, dense retrieval using semantic search with Sentence Transformers, and a proposed hybrid retrieval module. The hybrid module incorporates query expansion and combines the results of sparse and dense retrievers through a dynamically weighted Reciprocal Rank Fusion score. Using the HaluBench dataset, a benchmark for hallucinations in question answering tasks, we assess retrieval performance with metrics such as mean average precision and normalised discounted cumulative gain, focusing on the relevance of the top three retrieved documents. Results show that the hybrid retriever achieves better relevance scores, outperforming both sparse and dense retrievers. Further evaluation of LLM-generated answers against ground truth using metrics such as accuracy, hallucination rate, and rejection rate reveals that the hybrid retriever achieves the highest accuracy on fails, the lowest hallucination rate, and the lowest rejection rate. These findings highlight the hybrid retriever's ability to enhance retrieval relevance, reduce hallucination rates, and improve LLM reliability, emphasising the importance of advanced retrieval techniques in mitigating hallucinations and improving response accuracy.

cs.IR↗

arXiv2023

Keywords for Bias

This work proposes to analyse some keywords for bias analysis. For this, we are using several NLP approaches and compare them based on their capability of detecting keywords to analyse bias. The overall findings show that our proposed approach gives comparable results with the state-of-the-art approaches on different benchmark datasets.

cs.CY↗

arXiv2023

Experimenting with Selected Automated Approaches for Bias Analysis

This work first presents our attempts to establish an automated model using state-of-the-art approaches for analysing bias in search results of Bing and Google. Experimental results indicate that the current class-wise F1-scores of our best model are not sufficient to establish an automated model for bias analysis. Thus, we decided not to continue with this approach.

cs.IR↗

arXiv2023

Evaluation Metrics for Measuring Bias in Search Engine Results

Search engines decide what we see for a given search query. Since many people are exposed to information through search engines, it is fair to expect that search engines are neutral. However, search engine results do not necessarily cover all the viewpoints of a search query topic, and they can be biased towards a specific view since search engine results are returned based on relevance, which is calculated using many features and sophisticated algorithms where search neutrality is not necessarily the focal point. Therefore, it is important to evaluate the search engine results with respect to bias. In this work we propose novel web search bias evaluation measures which take into account the rank and relevance. We also propose a framework to evaluate web search bias using the proposed measures and test our framework on two popular search engines based on 57 controversial query topics such as abortion, medical marijuana, and gay marriage. We measure the stance bias (in support or against), as well as the ideological bias (conservative or liberal). We observe that the stance does not necessarily correlate with the ideological leaning, e.g. a positive stance on abortion indicates a liberal leaning but a positive stance on Cuba embargo indicates a conservative leaning. Our experiments show that neither of the search engines suffers from stance bias. However, both search engines suffer from ideological bias, both favouring one ideological leaning to the other, which is more significant from the perspective of polarisation in our society.

cs.IR↗

arXiv2022

Preliminary Bias Results in Search Engines

This report aims to report my thesis progress so far. My work attempts to show the differences in the perspectives of two search engines, Bing and Google on several selected controversial topics. In this work, we try to make a distinction on the viewpoints of Bing \& Google by using sentiment as well as the ranking of the document returned from these two search engines on the same queries, these queries are related mainly to controversial topics. You can find the methods we used with experimental results below.

cs.IR↗

arXiv2022

Quantifying Political Bias in News Articles

Search bias analysis is getting more attention in recent years since search results could affect In this work, we aim to establish an automated model for evaluating ideological bias in online news articles. The dataset is composed of news articles in search results as well as the newspaper articles. The current automated model results show that model capability is not sufficient to be exploited for annotating the documents automatically, thereby computing bias in search results.

cs.IR↗

arXiv2022

Case Study: The Impact of Location on Bias in Search Results

In this work, we aim to investigate the impact of location (different countries) on bias in search results. For this, we use the search results of Google and Bing in the UK and US locations. The query set is composed of controversial queries obtained from ProCon.org that have specific ideological leanings as conservative or liberal. In a previous work, researchers analyse search results in terms of stance and ideological bias with rank and relevance based measures. Yet, in the scope of this work, by using the query subset of controversial queries we examine the effect of location on the existence of bias as well as the magnitude of bias difference between Bing and Google. Note that this study follows a similar evaluation procedure. Our preliminary results show that location might affect the retrieval performance of search engines as well as the bias in the search results returned by Bing and Google towards the controversial queries.

cs.IR↗

arXiv2022

Measuring Gender Bias in Educational Videos: A Case Study on YouTube

Students are increasingly using online materials to learn new subjects or to supplement their learning process in educational institutions. Issues regarding gender bias have been raised in the context of formal education and some measures have been proposed to mitigate them. However, online educational materials in terms of possible gender bias and stereotypes which may appear in different forms are yet to be investigated in the context of search bias in a widely-used search platform. As a first step towards measuring possible gender bias in online platforms, we have investigated YouTube educational videos in terms of the perceived gender of their narrators. We adopted bias measures for ranked search results to evaluate educational videos returned by YouTube in response to queries related to STEM (Science, Technology, Engineering, and Mathematics) and NON-STEM fields of education. Gender is a research area by itself in social sciences which is beyond the scope of this work. In this respect, for annotating the perceived gender of the narrator of an instructional video we used only a crude classification of gender into Male, and Female. Then, for analysing perceived gender bias we utilised bias measures that have been inspired by search platforms and further incorporated rank information into our analysis. Our preliminary results demonstrate that there is a significant bias towards the male gender on the returned YouTube educational videos, and the degree of bias varies when we compare STEM and NON-STEM queries. Finally, there is a strong evidence that rank information might affect the results.

cs.IR↗

arXiv2021

Literature Review of the Pioneering Approaches in Cloud-based Search Engines Powered by LETOR Techniques

Search engines play an essential role in our daily lives. Nonetheless, they are also very crucial in enterprise domain to access documents from various information sources. Since traditional search systems index the documents mainly by looking at the frequency of the occurring words in these documents, they are barely able to support natural language search, but rather keyword search. It seems that keyword based search will not be sufficient for enterprise data which is growing extremely fast. Thus, enterprise search becomes increasingly critical in corporate domain. In this report, we present an overview of the state-of-the-art technologies in literature for three main purposes: i) to increase the retrieval performance of a search engine, ii) to deploy a search platform to a cloud environment, and iii) to select the best terms in expanding queries for achieving even a higher retrieval performance as well as to provide good query suggestions to its users for a better user experience.

cs.IR↗

arXiv2021

Query Suggestion for Click-Absent Queries in Enterprise Search

Creating alternative queries, also known as query suggestion, has been proved to be helpful on improving users' search experience. Owing to the suggestions, users could retrieve their information need more quickly and accurately. In many scenarios, these suggestions could be generated from the click-through logs by establishing a bipartite graph of the clicked query-document pairs. Most of the existing methods focused on click-existing queries which possess clicked information in the search logs, to suggest related queries using the co-clicked documents. In this paper, we propose a simple yet effective query suggestion method particularly for click-absent queries by ensuring semantic consistency without utilising any additional resources. Our experimental results show that the proposed technique generates comparatively good suggestions for click-absent queries on a real bilingual enterprise search log.

cs.IR↗