Sina Ahmadi, Rico Sennrich, Erfan Karami, Ako Marani, Parviz Fekrazad, Gholamreza Akbarzadeh Baghban, Hanah Hadi, Semko Heidari, Mahîr Dogan, Pedram Asadi, Dashne Bashir, Mohammad Amin Ghodrati, Kourosh Amini, Zeynab Ashourinezhad, Mana Baladi, Farshid Ezzati, Alireza Ghasemifar, Daryoush Hosseinpour, Behrooz Abbaszadeh, Amin Hassanpour, Bahaddin Jalal Hamaamin, Saya Kamal Hama, Ardeshir Mousavi, Sarko Nazir Hussein, Isar Nejadgholi, Mehmet Ölmez, Horam Osmanpour, Rashid Roshan Ramezani, Aryan Sediq Aziz, Ali Salehi, Mohammadreza Yadegari, Kewyar Yadegari, Sedighe Zamani Roodsari. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

While AI has been frequently applied in the context of immigration, most of these applications focus on selection and screening, which primarily serve to empower states and authorities, raising concerns due to their understudied reliability and high impact on immigrants' lives. In contrast, this paper emphasizes the potential of AI in Canada’s immigration settlement phase, a stage where access to information is crucial and service providers are overburdened. By highlighting the settlement sector as a prime candidate for reliable AI applications, we demonstrate its unique capacity to empower immigrants directly, yet it remains under-explored in AI research. We outline a vision for human-centred and responsible AI solutions that facilitate the integration of newcomers. We call on AI researchers to build upon our work and engage in multidisciplinary research and active collaboration with service providers and government organizations to develop tailored AI tools that are empowering, inclusive and safe.

Mitigation of gender bias in NLP has a long history tied to debiasing static word embeddings. More recently, attention has shifted to debiasing pre-trained language models. We study to what extent the simplest projective debiasing methods, developed for word embeddings, can help when applied to BERT’s internal representations. Projective methods are fast to implement, use a small number of saved parameters, and make no updates to the existing model parameters. We evaluate the efficacy of the methods in reducing both intrinsic bias, as measured by BERT’s next sentence prediction task, and in mitigating observed bias in a downstream setting when fine-tuned. To this end, we also provide a critical analysis of a popular gender-bias assessment test for quantifying intrinsic bias, resulting in an enhanced test set and new bias measures. We find that projective methods can be effective at both intrinsic bias and downstream bias mitigation, but that the two outcomes are not necessarily correlated. This finding serves as a warning that intrinsic bias test sets, based either on language modeling tasks or next sentence prediction, should not be the only benchmark in developing a debiased language model.

Gender stereotypes are pervasive beliefs about individuals based on their gender that play a significant role in shaping societal attitudes, behaviours, and even opportunities. Recognizing the negative implications of gender stereotypes, particularly in online communications, this study investigates eleven strategies to automatically counteract and challenge these views. We present AI-generated gender-based counter-stereotypes to (self-identified) male and female study participants and ask them to assess their offensiveness, plausibility, and potential effectiveness. The strategies of counter-facts and broadening universals (i.e., stating that anyone can have a trait regardless of group membership) emerged as the most robust approaches, while humour, perspective-taking, counter-examples, and empathy for the speaker were perceived as less effective. Also, the differences in ratings were more pronounced for stereotypes about the different targets than between the genders of the raters. Alarmingly, many AI-generated counter-stereotypes were perceived as offensive and/or implausible. Our analysis and the collected dataset offer foundational insight into counter-stereotype generation, guiding future efforts to develop strategies that effectively challenge gender stereotypes in online interactions.

This work provides an explanatory view of how LLMs can apply moral reasoning to both criticize and defend sexist language.We assessed eight large language models, all of which demonstrated the capability to provide explanations grounded in varying moral perspectives for both critiquing and endorsing views that reflect sexist assumptions.With both human and automatic evaluation, we show that all eight models produce comprehensible and contextually relevant text, which is helpful in understanding diverse views on how sexism is perceived.Also, through analysis of moral foundations cited by LLMs in their arguments, we uncover the diverse ideological perspectives in models' outputs, with some models aligning more with progressive or conservative views on gender roles and sexism.Based on our observations, we caution against the potential misuse of LLMs to justify sexist language.We also highlight that LLMs can serve as tools for understanding the roots of sexist beliefs and designing well-informed interventions.Given this dual capacity, it is crucial to monitor LLMs and design safety mechanisms for their use in applications that involve sensitive societal topics, such as sexism.Warning: This paper includes examples that might be offensive and upsetting.

Robustness of machine learning models on ever-changing real-world data is critical, especially for applications affecting human wellbeing such as content moderation.New kinds of abusive language continually emerge in online discussions in response to current events (e.g., COVID-19), and the deployed abuse detection systems should be updated regularly to remain accurate.In this paper, we show that general abusive language classifiers tend to be fairly reliable in detecting out-of-domain explicitly abusive utterances but fail to detect new types of more subtle, implicit abuse.Next, we propose an interpretability technique, based on the Testing Concept Activation Vector (TCAV) method from computer vision, to quantify the sensitivity of a trained model to the humandefined concepts of explicit and implicit abusive language, and use that to explain the generalizability of the model on new data, in this case, COVID-related anti-Asian hate speech.Extending this technique, we introduce a novel metric, Degree of Explicitness, for a single instance and show that the new metric is beneficial in suggesting out-of-domain unlabeled examples to effectively enrich the training data with informative, implicitly abusive texts.

Age-related stereotypes are pervasive in our society, and yet have been under-studied in the NLP community. Here, we present a method for extracting age-related stereotypes from Twitter data, generating a corpus of 300,000 over-generalizations about four contemporary generations (baby boomers, generation X, millennials, and generation Z), as well as “old” and “young” people more generally. By employing word-association metrics, semi-supervised topic modelling, and density-based clustering, we uncover many common stereotypes as reported in the media and in the psychological literature, as well as some more novel findings. We also observe trends consistent with the existing literature, namely that definitions of “young” and “old” age appear to be context-dependent, stereotypes for different generations vary across different topics (e.g., work versus family life), and some age-based stereotypes are distinct from generational stereotypes. The method easily extends to other social group labels, and therefore can be used in future work to study stereotypes of different social categories. By better understanding how stereotypes are formed and spread, and by tracking emerging stereotypes, we hope to eventually develop mitigating measures against such biased statements.

Esma Balkir, Isar Nejadgholi, Kathleen Fraser, Svetlana Kiritchenko. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

Kathleen C. Fraser, Isar Nejadgholi, Svetlana Kiritchenko. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

The pervasiveness of abusive content on the internet can lead to severe psychological and physical harm. Significant effort in Natural Language Processing (NLP) research has been devoted to addressing this problem through abusive content detection and related sub-areas, such as the detection of hate speech, toxicity, cyberbullying, etc. Although current technologies achieve high classification performance in research studies, it has been observed that the real-life application of this technology can cause unintended harms, such as the silencing of under-represented groups. We review a large body of NLP research on automatic abuse detection with a new focus on ethical challenges, organized around eight established ethical principles: privacy, accountability, safety and security, transparency and explainability, fairness and non-discrimination, human control of technology, professional responsibility, and promotion of human values. In many cases, these principles relate not only to situational ethical codes, which may be context-dependent, but are in fact connected to universal human rights, such as the right to privacy, freedom from discrimination, and freedom of expression. We highlight the need to examine the broad social impacts of this technology, and to bring ethical and human rights considerations to every stage of the application life-cycle, from task formulation and dataset design, to model training and evaluation, to application deployment. Guided by these principles, we identify several opportunities for rights-respecting, socio-technical solutions to detect and confront online abuse, including ‘nudging’, ‘quarantining’, value sensitive design, counter-narratives, style transfer, and AI-driven public education applications.evaluation, to application deployment. Guided by these principles, we identify several opportunities for rights-respecting, socio-technical solutions to detect and confront online abuse, including 'nudging', 'quarantining', value sensitive design, counter-narratives, style transfer, and AI-driven public education applications.

Classification is presented as a pre‐processing step in this study. The state of the subject is classified as the unmoving state with normal breathing (normal breathing class), unmoving state with no breathing (stop breathing class) or the state when the subject is moving (erratic signal class) before breathing estimation algorithms are applied. Estimation algorithms may be applied to obtain breathing rate if normal breathing class is detected or alarms may be generated if stop breathing is detected, and fine‐grained classification of activities may be pursued if the erratic signal is detected. Experiments were performed using a single‐channel pulse‐modulated continuous wave radar with three subjects for a total of 135 min. In each experiment, the subject was continuously monitored for 15 min and the subject performed activities that resulted in a signal that belonged to one of the three classes. Besides extracting a feature that assessed the distribution of energy of the signal in the frequency domain, a novel nonlinear time series feature extraction method based on the higher‐dimensional embedding technique was applied to ascertain periodicity of the reflected signal. Bayes classifier was used to classify each 5‐s segment of radar returns. A 30‐fold cross validation resulted in 97% of overall classification accuracy.

A semi-supervised approach was introduced to develop a semantic search system, capable of finding legal cases whose fact-asserting sentences are similar to a given query, in a large legal corpus. First, an unsupervised word embedding model learns the meaning of legal words from a large immigration law corpus. Then this knowledge is used to initiate the training of a fact detecting classifier with a small set of annotated legal cases. We achieved 90% accuracy in detecting fact sentences, where only 150 annotated documents were available. The hidden layer of the trained classifier is used to vectorize sentences and calculate cosine similarity between fact-asserting sentences and the given queries. We reached 78% mean average precision score in searching semantically similar sentences.

This paper presents a novel algorithm for radar-based estimation of vital signs in a noncontact, privacy friendly manner while subjects are in motion. Unlike the traditional methods that merely use the Fourier spectrum of the output of the radar receiver to obtain estimates of breathing and heart rates, the proposed algorithm uses time-frequency approach. From the Time-Frequency Representation (TFR) of the output of a pseudo-random binary Phase Modulated Continuous Wave (PMCW) radar, frequency of the maximum amplitude at every time instant is estimated and a timeseries of dominant frequencies is formed. MUSIC algorithm is then applied to estimate the vital signs from this series. The proposed algorithm is demonstrated using simulated and real data. Simulated data is obtained through modeling the output of a PMCW radar. Real data is obtained by monitoring a walking subject for 10 minutes in a realistic setting with a 24.125 GHz PMCW radar. The vital sign estimates obtained using the proposed method are found to match closely the estimates from wearable devices that were applied to provide the ground truth for breathing and heart rates.

According to some biological observations, generating output variability is one of the characteristics expected from a memory model. In this paper a BAM inspired chaotic model is used to mimic this functionality of the brain. Chaos gives the potential to create deterministic variability and control its degree of uncertainty. Using some time series generated by the trained network, largest lyapunov exponent is computed for different values of transient parameter of neurons' activation functions. Critical values of this parameter leading to most chaotic behavior for each neuron are stored and used to set the network during recall. Exhibiting desired behaviors with various degrees of uncertainty is achieved as a product of a complex interaction between a group of chaotic neurons and another group behaving in a fixed point manner.