Hossein A. Rahmani

Hossein A. Rahmani, Mohammadmehdi Naghiaei, Mahdi Dehghan et al.

Recent work in recommender systems mainly focuses on fairness in recommendations as an important aspect of measuring recommendations quality. A fairness-aware recommender system aims to treat different user groups similarly. Relevant work on user-oriented fairness highlights the discriminative behavior of fairness-unaware recommendation algorithms towards a certain user group, defined based on users' activity level. Typical solutions include proposing a user-centered fairness re-ranking framework applied on top of a base ranking model to mitigate its unfair behavior towards a certain user group i.e., disadvantaged group. In this paper, we re-produce a user-oriented fairness study and provide extensive experiments to analyze the dependency of their proposed method on various fairness and recommendation aspects, including the recommendation domain, nature of the base ranking model, and user grouping method. Moreover, we evaluate the final recommendations provided by the re-ranking framework from both user- (e.g., NDCG, user-fairness) and item-side (e.g., novelty, item-fairness) metrics. We discover interesting trends and trade-offs between the model's performance in terms of different evaluation metrics. For instance, we see that the definition of the advantaged/disadvantaged user groups plays a crucial role in the effectiveness of the fairness algorithm and how it improves the performance of specific base ranking models. Finally, we highlight some important open challenges and future directions in this field. We release the data, evaluation pipeline, and the trained models publicly on https://github.com/rahmanidashti/FairRecSys.

Mubashara Akhtar, Anka Reuel, Prajna Soni et al. · meta-ai

Artificial Intelligence (AI) benchmarks play a central role in measuring progress in model development and guiding deployment decisions. However, many benchmarks quickly become saturated, meaning that they can no longer differentiate between the best-performing models, diminishing their long-term value. In this study, we analyze benchmark saturation across 60 Large Language Model (LLM) benchmarks selected from technical reports by major model developers. To identify factors driving saturation, we characterize benchmarks along 14 properties spanning task design, data construction, and evaluation format. We test five hypotheses examining how each property contributes to saturation rates. Our analysis reveals that nearly half of the benchmarks exhibit saturation, with rates increasing as benchmarks age. Notably, hiding test data (i.e., public vs. private) shows no protective effect, while expert-curated benchmarks resist saturation better than crowdsourced ones. Our findings highlight which design choices extend benchmark longevity and inform strategies for more durable evaluation.

Mohammadmehdi Naghiaei, Hossein A. Rahmani, Yashar Deldjoo

Recently, there has been a rising awareness that when machine learning (ML) algorithms are used to automate choices, they may treat/affect individuals unfairly, with legal, ethical, or economic consequences. Recommender systems are prominent examples of such ML systems that assist users in making high-stakes judgments. A common trend in the previous literature research on fairness in recommender systems is that the majority of works treat user and item fairness concerns separately, ignoring the fact that recommender systems operate in a two-sided marketplace. In this work, we present an optimization-based re-ranking approach that seamlessly integrates fairness constraints from both the consumer and producer-side in a joint objective framework. We demonstrate through large-scale experiments on 8 datasets that our proposed method is capable of improving both consumer and producer fairness without reducing overall recommendation quality, demonstrating the role algorithms may play in minimizing data biases.

Rikaz Rameez, Hossein A. Rahmani, Emine Yilmaz

Recently, Twitter has become the social network of choice for sharing and spreading information to a multitude of users through posts called 'tweets'. Users can easily re-share these posts to other users through 'retweets', which allow information to cascade to many more users, increasing its outreach. Clearly, being able to know the extent to which a post can be retweeted has great value in advertising, influencing and other such campaigns. In this paper we propose ViralBERT, which can be used to predict the virality of tweets using content- and user-based features. We employ a method of concatenating numerical features such as hashtags and follower numbers to tweet text, and utilise two BERT modules: one for semantic representation of the combined text and numerical features, and another module purely for sentiment analysis of text, as both the information within text and it's ability to elicit an emotional response play a part in retweet proneness. We collect a dataset of 330k tweets to train ViralBERT and validate the efficacy of our model using baselines from current studies in this field. Our experiments show that our approach outperforms these baselines, with a 13% increase in both F1 Score and Accuracy compared to the best performing baseline method. We then undergo an ablation study to investigate the importance of chosen features, finding that text sentiment and follower counts, and to a lesser extent mentions and following counts, are the strongest features for the model, and that hashtag counts are detrimental to the model.

Yinzhu Chen, Abdine Maiga, Hossein A. Rahmani et al.

Large Language Models (LLMs) are increasingly used for clinical decision support, where hallucinations and unsafe suggestions may pose direct risks to patient safety. These risks are particularly challenging as they often manifest as subtle clinical errors that evade detection by generic metrics, while expert-authored fine-grained rubrics remain costly to construct and difficult to scale. In this paper, we propose a retrieval-augmented multi-agent framework designed to automate the generation of instance-specific evaluation rubrics. Our approach grounds evaluation in authoritative medical evidence by decomposing retrieved content into atomic facts and synthesizing them with user interaction constraints to form verifiable, fine-grained evaluation criteria. Evaluated on HealthBench, our framework achieves a Clinical Intent Alignment (CIA) score of 60.12%, a statistically significant improvement over the GPT-4o baseline (55.16%). In discriminative tests, our rubrics yield a mean score delta ($μ_Δ = 8.658$) and an AUROC of 0.977, nearly doubling the quality separation achieved by GPT-4o baseline (4.972). Beyond evaluation, our rubrics effectively guide response refinement, improving quality by 9.2% (from 59.0% to 68.2%). This provides a scalable and transparent foundation for both evaluating and improving medical LLMs. The code is available at https://anonymous.4open.science/r/Automated-Rubric-Generation-AF3C/.

Xi Wang, Hossein A. Rahmani, Jiqun Liu et al.

Conversational Recommendation System (CRS) is a rapidly growing research area that has gained significant attention alongside advancements in language modelling techniques. However, the current state of conversational recommendation faces numerous challenges due to its relative novelty and limited existing contributions. In this study, we delve into benchmark datasets for developing CRS models and address potential biases arising from the feedback loop inherent in multi-turn interactions, including selection bias and multiple popularity bias variants. Drawing inspiration from the success of generative data via using language models and data augmentation techniques, we present two novel strategies, 'Once-Aug' and 'PopNudge', to enhance model performance while mitigating biases. Through extensive experiments on ReDial and TG-ReDial benchmark datasets, we show a consistent improvement of CRS techniques with our data augmentation approaches and offer additional insights on addressing multiple newly formulated biases.

Anka Reuel, Avijit Ghosh, Jenny Chim et al.

Foundation models are increasingly central to high-stakes AI systems, and governance frameworks now depend on evaluations to assess their risks and capabilities. Although general capability evaluations are widespread, social impact assessments covering bias, fairness, privacy, environmental costs, and labor practices remain uneven across the AI ecosystem. To characterize this landscape, we conduct the first comprehensive analysis of both first-party and third-party social impact evaluation reporting across a wide range of model developers. Our study examines 186 first-party release reports and 183 post-release evaluation sources, and complements this quantitative analysis with interviews of model developers. We find a clear division of evaluation labor: first-party reporting is sparse, often superficial, and has declined over time in key areas such as environmental impact and bias, while third-party evaluators including academic researchers, nonprofits, and independent organizations provide broader and more rigorous coverage of bias, harmful content, and performance disparities. However, this complementarity has limits. Only model developers can authoritatively report on data provenance, content moderation labor, financial costs, and training infrastructure, yet interviews reveal that these disclosures are often deprioritized unless tied to product adoption or regulatory compliance. Our findings indicate that current evaluation practices leave major gaps in assessing AI's societal impacts, highlighting the urgent need for policies that promote developer transparency, strengthen independent evaluation ecosystems, and create shared infrastructure to aggregate and compare third-party evaluations in a consistent and accessible way.

Jerome Ramos, Feng Xia, Xi Wang et al.

Training conversational recommender systems (CRS) requires extensive dialogue data, which is challenging to collect at scale. To address this, researchers have used simulated user-recommender conversations. Traditional simulation approaches often utilize a single large language model (LLM) that generates entire conversations with prior knowledge of the target items, leading to scripted and artificial dialogues. We propose a reference-free simulation framework that trains two independent LLMs, one as the user and one as the conversational recommender. These models interact in real-time without access to predetermined target items, but preference summaries and target attributes, enabling the recommender to genuinely infer user preferences through dialogue. This approach produces more realistic and diverse conversations that closely mirror authentic human-AI interactions. Our reference-free simulators match or exceed existing methods in quality, while offering a scalable solution for generating high-quality conversational recommendation data without constraining conversations to pre-defined target items. We conduct both quantitative and human evaluations to confirm the effectiveness of our reference-free approach.

Hossein A. Rahmani, Satyapriya Krishna, Xi Wang et al.

Large language models have recently demonstrated remarkable abilities to self-correct their responses through iterative refinement, often referred to as self-consistency or self-reflection. However, the dynamics of this self-correction mechanism may differ substantially depending on whether the model is tasked with open-ended text generation or with selecting the most appropriate response from multiple predefined options. In this paper, we conduct a systematic investigation of these two paradigms by comparing performance trends and error-correction behaviors across various natural language understanding and reasoning tasks, covering language models of different scales and families. Our experimental results reveal distinct patterns of improvement and failure modes: \textit{While open-ended generation often benefits from the flexibility of re-interpretation and compositional refinement, multiple-choice selection can leverage clearer solution boundaries but may be limited by the provided options}. This contrast also reflects the dual demands faced by emerging agentic LLM applications: effective agents must not only generate and refine open-ended plans or explanations, but also make reliable discrete choices when operating within constrained action spaces. Our findings, therefore, highlight that the design of self-correction mechanisms should take into account the interaction between task structure and output space, with implications for both knowledge-intensive reasoning and decision-oriented applications of LLMs.

Shuofei Qiao, Yunxiang Wei, Xuehai Wang et al.

The rapid evolution of Large Language Models has catalyzed a surge in scientific idea production, yet this leap has not been accompanied by a matching advance in idea evaluation. The fundamental nature of scientific evaluation needs knowledgeable grounding, collective deliberation, and multi-criteria decision-making. However, existing idea evaluation methods often suffer from narrow knowledge horizons, flattened evaluation dimensions, and the inherent bias in LLM-as-a-Judge. To address these, we regard idea evaluation as a knowledge-grounded, multi-perspective reasoning problem and introduce InnoEval, a deep innovation evaluation framework designed to emulate human-level idea assessment. We apply a heterogeneous deep knowledge search engine that retrieves and grounds dynamic evidence from diverse online sources. We further achieve review consensus with an innovation review board containing reviewers with distinct academic backgrounds, enabling a multi-dimensional decoupled evaluation across multiple metrics. We construct comprehensive datasets derived from authoritative peer-reviewed submissions to benchmark InnoEval. Experiments demonstrate that InnoEval can consistently outperform baselines in point-wise, pair-wise, and group-wise evaluation tasks, exhibiting judgment patterns and consensus highly aligned with human experts.

Hossein A. Rahmani, Varsha Ramineni, Emine Yilmaz et al.

Test collections are crucial for evaluating Information Retrieval (IR) systems. Creating a diverse set of user queries for these collections can be challenging, and obtaining relevance judgments, which indicate how well retrieved documents match a query, is often costly and resource-intensive. Recently, generating synthetic datasets using Large Language Models (LLMs) has gained attention in various applications. While previous work has used LLMs to generate synthetic queries or documents to improve ranking models, using LLMs to create synthetic test collections is still relatively unexplored. Previous work~\cite{rahmani2024synthetic} showed that synthetic test collections have the potential to be used for system evaluation, however, more analysis is needed to validate this claim. In this paper, we thoroughly investigate the reliability of synthetic test collections constructed using LLMs, where LLMs are used to generate synthetic queries, labels, or both. In particular, we examine the potential biases that might occur when such test collections are used for evaluation. We first empirically show the presence of such bias in evaluation results and analyse the effects it might have on system evaluation. We further validate the presence of such bias using a linear mixed-effects model. Our analysis shows that while the effect of bias present in evaluation results obtained using synthetic test collections could be significant, for e.g.~computing absolute system performance, its effect may not be as significant in comparing relative system performance. Codes and data are available at: https://github.com/rahmanidashti/BiasSyntheticData.

Yue Feng, Hossein A. Rahmani, Aldo Lipani et al.

Task-oriented dialogue systems aim at providing users with task-specific services. Users of such systems often do not know all the information about the task they are trying to accomplish, requiring them to seek information about the task. To provide accurate and personalized task-oriented information seeking results, task-oriented dialogue systems need to address two potential issues: 1) users' inability to describe their complex information needs in their requests; and 2) ambiguous/missing information the system has about the users. In this paper, we propose a new Multi-Attention Seq2Seq Network, named MAS2S, which can ask questions to clarify the user's information needs and the user's profile in task-oriented information seeking. We also extend an existing dataset for task-oriented information seeking, leading to the \ourdataset which contains about 100k task-oriented information seeking dialogues that are made publicly available\footnote{Dataset and code is available at \href{https://github.com/sweetalyssum/clarit}{https://github.com/sweetalyssum/clarit}.}. Experimental results on \ourdataset show that MAS2S outperforms baselines on both clarification question generation and answer prediction.

Hossein A. Rahmani, Nick Craswell, Emine Yilmaz et al.

Test collections play a vital role in evaluation of information retrieval (IR) systems. Obtaining a diverse set of user queries for test collection construction can be challenging, and acquiring relevance judgments, which indicate the appropriateness of retrieved documents to a query, is often costly and resource-intensive. Generating synthetic datasets using Large Language Models (LLMs) has recently gained significant attention in various applications. In IR, while previous work exploited the capabilities of LLMs to generate synthetic queries or documents to augment training data and improve the performance of ranking models, using LLMs for constructing synthetic test collections is relatively unexplored. Previous studies demonstrate that LLMs have the potential to generate synthetic relevance judgments for use in the evaluation of IR systems. In this paper, we comprehensively investigate whether it is possible to use LLMs to construct fully synthetic test collections by generating not only synthetic judgments but also synthetic queries. In particular, we analyse whether it is possible to construct reliable synthetic test collections and the potential risks of bias such test collections may exhibit towards LLM-based models. Our experiments indicate that using LLMs it is possible to construct synthetic test collections that can reliably be used for retrieval evaluation.

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al.

This is the fifth year of the TREC Deep Learning track. As in previous years, we leverage the MS MARCO datasets that made hundreds of thousands of human-annotated training labels available for both passage and document ranking tasks. We mostly repeated last year's design, to get another matching test set, based on the larger, cleaner, less-biased v2 passage and document set, with passage ranking as primary and document ranking as a secondary task (using labels inferred from passage). As we did last year, we sample from MS MARCO queries that were completely held out, unused in corpus construction, unlike the test queries in the first three years. This approach yields a more difficult test with more headroom for improvement. Alongside the usual MS MARCO (human) queries from MS MARCO, this year we generated synthetic queries using a fine-tuned T5 model and using a GPT-4 prompt. The new headline result this year is that runs using Large Language Model (LLM) prompting in some way outperformed runs that use the "nnlm" approach, which was the best approach in the previous four years. Since this is the last year of the track, future iterations of prompt-based ranking can happen in other tracks. Human relevance assessments were applied to all query types, not just human MS MARCO queries. Evaluation using synthetic queries gave similar results to human queries, with system ordering agreement of $τ=0.8487$. However, human effort was needed to select a subset of the synthetic queries that were usable. We did not see clear evidence of bias, where runs using GPT-4 were favored when evaluated using synthetic GPT-4 queries, or where runs using T5 were favored when evaluated on synthetic T5 queries.

Hossein A. Rahmani, Mohammadmehdi Naghiaei, Yashar Deldjoo

In recent years, there has been an increasing recognition that when machine learning (ML) algorithms are used to automate decisions, they may mistreat individuals or groups, with legal, ethical, or economic implications. Recommender systems are prominent examples of these machine learning (ML) systems that aid users in making decisions. The majority of past literature research on RS fairness treats user and item fairness concerns independently, ignoring the fact that recommender systems function in a two-sided marketplace. In this paper, we propose CP-FairRank, an optimization-based re-ranking algorithm that seamlessly integrates fairness constraints from both the consumer and producer side in a joint objective framework. The framework is generalizable and may take into account varied fairness settings based on group segmentation, recommendation model selection, and domain, which is one of its key characteristics. For instance, we demonstrate that the system may jointly increase consumer and producer fairness when (un)protected consumer groups are defined on the basis of their activity level and main-streamness, while producer groups are defined according to their popularity level. For empirical validation, through large-scale on eight datasets and four mainstream collaborative filtering (CF) recommendation models, we demonstrate that our proposed strategy is able to improve both consumer and producer fairness without compromising or very little overall recommendation quality, demonstrating the role algorithms may play in avoiding data biases.

Xiao Fu, Hossein A. Rahmani, Bin Wu et al.

Personalised text generation is essential for user-centric information systems, yet most evaluation methods overlook the individuality of users. We introduce \textbf{PREF}, a \textbf{P}ersonalised \textbf{R}eference-free \textbf{E}valuation \textbf{F}ramework that jointly measures general output quality and user-specific alignment without requiring gold personalised references. PREF operates in a three-step pipeline: (1) a coverage stage uses a large language model (LLM) to generate a comprehensive, query-specific guideline covering universal criteria such as factuality, coherence, and completeness; (2) a preference stage re-ranks and selectively augments these factors using the target user's profile, stated or inferred preferences, and context, producing a personalised evaluation rubric; and (3) a scoring stage applies an LLM judge to rate candidate answers against this rubric, ensuring baseline adequacy while capturing subjective priorities. This separation of coverage from preference improves robustness, transparency, and reusability, and allows smaller models to approximate the personalised quality of larger ones. Experiments on the PrefEval benchmark, including implicit preference-following tasks, show that PREF achieves higher accuracy, better calibration, and closer alignment with human judgments than strong baselines. By enabling scalable, interpretable, and user-aligned evaluation, PREF lays the groundwork for more reliable assessment and development of personalised language generation systems.

Lisa Alazraki, Lihu Chen, Ana Brassard et al.

Large Language Models (LLMs) have achieved high accuracy on complex commonsense and mathematical problems that involve the composition of multiple reasoning steps. However, current compositional benchmarks testing these skills tend to focus on either commonsense or math reasoning, whereas LLM agents solving real-world tasks would require a combination of both. In this work, we introduce an Agentic Commonsense and Math benchmark (AgentCoMa), where each compositional task requires a commonsense reasoning step and a math reasoning step. We test it on 61 LLMs of different sizes, model families, and training strategies. We find that LLMs can usually solve both steps in isolation, yet their accuracy drops by ~30% on average when the two are combined. This is a substantially greater performance gap than the one we observe in prior compositional benchmarks that combine multiple steps of the same reasoning type. In contrast, non-expert human annotators can solve the compositional questions and the individual steps in AgentCoMa with similarly high accuracy. Furthermore, we conduct a series of interpretability studies to better understand the performance gap, examining neuron patterns, attention maps and membership inference. Our work underscores a substantial degree of model brittleness in the context of mixed-type compositional reasoning and offers a test bed for future improvement.

Varsha Ramineni, Hossein A. Rahmani, Emine Yilmaz et al.

Ensuring fairness in AI systems is critical, especially in high-stakes domains such as lending, hiring, and healthcare. This urgency is reflected in emerging global regulations that mandate fairness assessments and independent bias audits. However, procuring the necessary complete data for fairness testing remains a significant challenge. In industry settings, legal and privacy concerns restrict the collection of demographic data required to assess group disparities, and auditors face practical and cultural challenges in gaining access to data. In practice, data relevant for fairness testing is often split across separate sources: internal datasets held by institutions with predictive attributes, and external public datasets such as census data containing protected attributes, each providing only partial, marginal information. Our work seeks to leverage such available separate data to estimate model fairness when complete data is inaccessible. We propose utilising the available separate data to estimate a set of feasible joint distributions and then compute the set plausible fairness metrics. Through simulation and real experiments, we demonstrate that we can derive meaningful bounds on fairness metrics and obtain reliable estimates of the true metric. Our results demonstrate that this approach can serve as a practical and effective solution for fairness testing in real-world settings where access to complete data is restricted.

Varsha Ramineni, Hossein A. Rahmani, Emine Yilmaz et al.

As AI becomes prevalent in high-risk domains and decision-making, it is essential to test for potential harms and biases. This urgency is reflected by the global emergence of AI regulations that emphasise fairness and adequate testing, with some mandating independent bias audits. However, procuring the necessary data for fairness testing remains a significant challenge. Particularly in industry settings, legal and privacy concerns restrict the collection of demographic data required to assess group disparities, and auditors face practical and cultural challenges in gaining access to data. Further, internal historical datasets are often insufficiently representative to identify real-world biases. This work focuses on evaluating classifier fairness when complete datasets including demographics are inaccessible. We propose leveraging separate overlapping datasets to construct complete synthetic data that includes demographic information and accurately reflects the underlying relationships between protected attributes and model features. We validate the fidelity of the synthetic data by comparing it to real data, and empirically demonstrate that fairness metrics derived from testing on such synthetic data are consistent with those obtained from real data. This work, therefore, offers a path to overcome real-world data scarcity for fairness testing, enabling independent, model-agnostic evaluation of fairness, and serving as a viable substitute where real data is limited.

Bin Wu, Zhengyan Shi, Hossein A. Rahmani et al.

Utilizing user profiles to personalize Large Language Models (LLMs) has been shown to enhance the performance on a wide range of tasks. However, the precise role of user profiles and their effect mechanism on LLMs remains unclear. This study first confirms that the effectiveness of user profiles is primarily due to personalization information rather than semantic information. Furthermore, we investigate how user profiles affect the personalization of LLMs. Within the user profile, we reveal that it is the historical personalized response produced or approved by users that plays a pivotal role in personalizing LLMs. This discovery unlocks the potential of LLMs to incorporate a greater number of user profiles within the constraints of limited input length. As for the position of user profiles, we observe that user profiles integrated into different positions of the input context do not contribute equally to personalization. Instead, where the user profile that is closer to the beginning affects more on the personalization of LLMs. Our findings reveal the role of user profiles for the personalization of LLMs, and showcase how incorporating user profiles impacts performance providing insight to leverage user profiles effectively.

Zhengxiang Shi, Jerome Ramos, To Eun Kim et al.

In collaborative tasks, effective communication is crucial for achieving joint goals. One such task is collaborative building where builders must communicate with each other to construct desired structures in a simulated environment such as Minecraft. We aim to develop an intelligent builder agent to build structures based on user input through dialogue. However, in collaborative building, builders may encounter situations that are difficult to interpret based on the available information and instructions, leading to ambiguity. In the NeurIPS 2022 Competition NLP Task, we address two key research questions, with the goal of filling this gap: when should the agent ask for clarification, and what clarification questions should it ask? We move towards this target with two sub-tasks, a classification task and a ranking task. For the classification task, the goal is to determine whether the agent should ask for clarification based on the current world state and dialogue history. For the ranking task, the goal is to rank the relevant clarification questions from a pool of candidates. In this report, we briefly introduce our methods for the classification and ranking task. For the classification task, our model achieves an F1 score of 0.757, which placed the 3rd on the leaderboard. For the ranking task, our model achieves about 0.38 for Mean Reciprocal Rank by extending the traditional ranking model. Lastly, we discuss various neural approaches for the ranking task and future direction.

Mohammadmehdi Naghiaei, Hossein A. Rahmani, Mahdi Dehghan

Recent studies have shown that recommendation systems commonly suffer from popularity bias. Popularity bias refers to the problem that popular items (i.e., frequently rated items) are recommended frequently while less popular items are recommended rarely or not at all. Researchers adopted two approaches to examining popularity bias: (i) from the users' perspective, by analyzing how far a recommendation system deviates from user's expectations in receiving popular items, and (ii) by analyzing the amount of exposure that long-tail items receive, measured by overall catalog coverage and novelty. In this paper, we examine the first point of view in the book domain, although the findings may be applied to other domains as well. To this end, we analyze the well-known Book-Crossing dataset and define three user groups based on their tendency towards popular items (i.e., Niche, Diverse, Bestseller-focused). Further, we evaluate the performance of nine state-of-the-art recommendation algorithms and two baselines (i.e., Random, MostPop) from both the accuracy (e.g., NDCG, Precision, Recall) and popularity bias perspectives. Our results indicate that most state-of-the-art recommendation algorithms suffer from popularity bias in the book domain, and fail to meet users' expectations with Niche and Diverse tastes despite having a larger profile size. Conversely, Bestseller-focused users are more likely to receive high-quality recommendations, both in terms of fairness and personalization. Furthermore, our study shows a tradeoff between personalization and unfairness of popularity bias in recommendation algorithms for users belonging to the Diverse and Bestseller groups, that is, algorithms with high capability of personalization suffer from the unfairness of popularity bias.

Hossein A. Rahmani, Yashar Deldjoo, Ali Tourani et al.

Point-of-Interest (POI) recommender systems provide personalized recommendations to users and help businesses attract potential customers. Despite their success, recent studies suggest that highly data-driven recommendations could be impacted by data biases, resulting in unfair outcomes for different stakeholders, mainly consumers (users) and providers (items). Most existing fairness-related research works in recommender systems treat user fairness and item fairness issues individually, disregarding that RS work in a two-sided marketplace. This paper studies the interplay between (i) the unfairness of active users, (ii) the unfairness of popular items, and (iii) the accuracy (personalization) of recommendation as three angles of our study triangle. We group users into advantaged and disadvantaged levels to measure user fairness based on their activity level. For item fairness, we divide items into short-head, mid-tail, and long-tail groups and study the exposure of these item groups into the top-k recommendation list of users. Experimental validation of eight different recommendation models commonly used for POI recommendation (e.g., contextual, CF) on two publicly available POI recommendation datasets, Gowalla and Yelp, indicate that most well-performing models suffer seriously from the unfairness of popularity bias (provider unfairness). Furthermore, our study shows that most recommendation models cannot satisfy both consumer and producer fairness, indicating a trade-off between these variables possibly due to natural biases in data. We choose the POI recommendation as our test scenario; however, the insights should be trivially extendable on other domains.

Hossein A. Rahmani, Mohammad Aliannejadi, Mitra Baratchi et al.

As the popularity of Location-based Social Networks (LBSNs) increases, designing accurate models for Point-of-Interest (POI) recommendation receives more attention. POI recommendation is often performed by incorporating contextual information into previously designed recommendation algorithms. Some of the major contextual information that has been considered in POI recommendation are the location attributes (i.e., exact coordinates of a location, category, and check-in time), the user attributes (i.e., comments, reviews, tips, and check-in made to the locations), and other information, such as the distance of the POI from user's main activity location, and the social tie between users. The right selection of such factors can significantly impact the performance of the POI recommendation. However, previous research does not consider the impact of the combination of these different factors. In this paper, we propose different contextual models and analyze the fusion of different major contextual information in POI recommendation. The major contributions of this paper are: (i) providing an extensive survey of context-aware location recommendation (ii) quantifying and analyzing the impact of different contextual information (e.g., social, temporal, spatial, and categorical) in the POI recommendation on available baselines and two new linear and non-linear models, that can incorporate all the major contextual information into a single recommendation model, and (iii) evaluating the considered models using two well-known real-world datasets. Our results indicate that while modeling geographical and temporal influences can improve recommendation quality, fusing all other contextual information into a recommendation model is not always the best strategy.

Kosar Seyedhoseinzadeh, Hossein A. Rahmani, Mohsen Afsharchi et al.

Recommender Systems (RSs) aim to model and predict the user preference while interacting with items, such as Points of Interest (POIs). These systems face several challenges, such as data sparsity, limiting their effectiveness. In this paper, we address this problem by incorporating social, geographical, and temporal information into the Matrix Factorization (MF) technique. To this end, we model social influence based on two factors: similarities between users in terms of common check-ins and the friendships between them. We introduce two levels of friendship based on explicit friendship networks and high check-in overlap between users. We base our friendship algorithm on users' geographical activity centers. The results show that our proposed model outperforms the state-of-the-art on two real-world datasets. More specifically, our ablation study shows that the social model improves the performance of our proposed POI recommendation system by 31% and 14% on the Gowalla and Yelp datasets in terms of Precision@10, respectively.

Hossein A. Rahmani, Jie Yang

Key Opinion Leaders (KOLs) are people that have a strong influence and their opinions are listened to by people when making important decisions. Crowdsourcing provides an efficient and cost-effective means to gather data for the KOL finding task. However, data collected through crowdsourcing is affected by the inherent demographic biases of crowd workers. To avoid such demographic biases, we need to measure how biased each crowd worker is. In this paper, we propose a simple yet effective approach based on demographic information of candidate KOLs and their counterfactual value. We argue that it is effectiveness because of the extra information that we can consider together with labeled data to curate a less biased dataset.

Mahdi Dehghan, Hossein A. Rahmani, Ahmad Ali Abin et al.

Expert finding addresses the task of retrieving and ranking talented people on the subject of user query. It is a practical issue in the Community Question Answering networks. Recruiters looking for knowledgeable people for their job positions are the most important clients of expert finding systems. In addition to employee expertise, the cost of hiring new staff is another significant concern for organizations. An efficient solution to cope with this concern is to hire T-shaped experts that are cost-effective. In this study, we have proposed a new deep model for T-shaped experts finding based on Convolutional Neural Networks. The proposed model tries to match queries and users by extracting local and position-invariant features from their corresponding documents. In other words, it detects users' shape of expertise by learning patterns from documents of users and queries simultaneously. The proposed model contains two parallel CNN's that extract latent vectors of users and queries based on their corresponding documents and join them together in the last layer to match queries with users. Experiments on a large subset of Stack Overflow documents indicate the effectiveness of the proposed method against baselines in terms of NDCG, MRR, and ERR evaluation metrics.

Hossein A. Rahmani, Mohammad Aliannejadi, Mitra Baratchi et al.

With the popularity of Location-based Social Networks, Point-of-Interest (POI) recommendation has become an important task, which learns the users' preferences and mobility patterns to recommend POIs. Previous studies show that incorporating contextual information such as geographical and temporal influences is necessary to improve POI recommendation by addressing the data sparsity problem. However, existing methods model the geographical influence based on the physical distance between POIs and users, while ignoring the temporal characteristics of such geographical influences. In this paper, we perform a study on the user mobility patterns where we find out that users' check-ins happen around several centers depending on their current temporal state. Next, we propose a spatio-temporal activity-centers algorithm to model users' behavior more accurately. Finally, we demonstrate the effectiveness of our proposed contextual model by incorporating it into the matrix factorization model under two different settings: i) static and ii) temporal. To show the effectiveness of our proposed method, which we refer to as STACP, we conduct experiments on two well-known real-world datasets acquired from Gowalla and Foursquare LBSNs. Experimental results show that the STACP model achieves a statistically significant performance improvement, compared to the state-of-the-art techniques. Also, we demonstrate the effectiveness of capturing geographical and temporal information for modeling users' activity centers and the importance of modeling them jointly.

Hossein A. Rahmani, Mohammad Aliannejadi, Sajad Ahmadian et al.

With the rapid growth of Location-Based Social Networks, personalized Points of Interest (POIs) recommendation has become a critical task to help users explore their surroundings. Due to the scarcity of check-in data, the availability of geographical information offers an opportunity to improve the accuracy of POI recommendation. Moreover, matrix factorization methods provide effective models which can be used in POI recommendation. However, there are two main challenges which should be addressed to improve the performance of POI recommendation methods. First, leveraging geographical information to capture both the user's personal, geographic profile and a location's geographic popularity. Second, incorporating the geographical model into the matrix factorization approaches. To address these problems, a POI recommendation method is proposed in this paper based on a Local Geographical Model, which considers both users' and locations' points of view. To this end, an effective geographical model is proposed by considering the user's main region of activity and the relevance of each location within that region. Then, the proposed local geographical model is fused into the Logistic Matrix Factorization to improve the accuracy of POI recommendation. Experimental results on two well-known datasets demonstrate that the proposed approach outperforms other state-of-the-art POI recommendation methods.

Hossein A. Rahmani, Mohammad Aliannejadi, Rasoul Mirzaei Zadeh et al.

Recently, Point of interest (POI) recommendation has gained ever-increasing importance in various Location-Based Social Networks (LBSNs). With the recent advances of neural models, much work has sought to leverage neural networks to learn neural embeddings in a pre-training phase that achieve an improved representation of POIs and consequently a better recommendation. However, previous studies fail to capture crucial information about POIs such as categorical information. In this paper, we propose a novel neural model that generates a POI embedding incorporating sequential and categorical information from POIs. Our model consists of a check-in module and a category module. The check-in module captures the geographical influence of POIs derived from the sequence of users' check-ins, while the category module captures the characteristics of POIs derived from the category information. To validate the efficacy of the model, we experimented with two large-scale LBSN datasets. Our experimental results demonstrate that our approach significantly outperforms state-of-the-art POI recommendation methods.