arXiv subjects

David Lillis

Publications and source records attributed to David Lillis.

At least 19 recordsLinked to original sources

arXiv2026

Assessing and Explaining the Persuadability of Large Language Models as Legal Decision Tools

As Large Language Models (LLMs) are proposed as legal decision assistants, and even first-instance decision-makers, across a range of judicial and administrative contexts, it becomes essential to explore how they answer legal questions, and in particular the factors that lead them to decide difficult questions. A specific feature of legal decisions is the need to respond to arguments advanced by contending parties. A legal decision-maker must be able to engage with, and respond to, including through being potentially persuaded by, these arguments. Conversely, they should not be unduly persuadable, deciding cases based on the skills of the advocates rather than the merits of the case. In this paper we explore how frontier open- and closed-weights LLMs respond to legal arguments. We propose a metric to measure persuadability in the trilateral setting in which competing advocates seek to persuade a judge of opposite conclusions. We report original experimental results measuring how far the quality of the advocate making arguments affects the likelihood that a given model will agree with a particular legal point of view. We further examine how far models are capable of distinguishing between stronger and weaker arguments and how far model judgments in this domain are affected by positional bias. Through parallel bilateral trials we show how the trilateral setting changes the demands on judge models, and in turn their apparent persuadability. Finally we examine the specific features of arguments that affect persuasion, including the relative contribution of legal content and rhetorical form, the extent to which model persuasion tracks human expert judgments of argument quality, and the extent to which argument quantity, diversity and type affect persuasive outcomes. Our results have implications for the feasibility of adopting LLMs across legal and administrative settings.

cs.AI↗

arXiv2026

Why AI-Generated Text Detection Fails: Evidence from Explainable AI Beyond Benchmark Accuracy

The widespread adoption of Large Language Models (LLMs) has made the detection of AI-Generated text a pressing and complex challenge. Although many detection systems report high benchmark accuracy, their reliability in real-world settings remains uncertain, and their interpretability is often unexplored. In this work, we investigate whether contemporary detectors genuinely identify machine authorship or merely exploit dataset-specific artefacts. We propose an interpretable detection framework that integrates linguistic feature engineering, machine learning, and explainable AI techniques. When evaluated on two prominent benchmark corpora, namely PAN CLEF 2025 and COLING 2025, our model trained on 30 linguistic features achieves leaderboard-competitive performance, attaining an F1 score of 0.9734. However, systematic cross-domain and cross-generator evaluation reveals substantial generalisation failure: classifiers that excel in-domain degrade significantly under distribution shift. Using SHAP- based explanations, we show that the most influential features differ markedly between datasets, indicating that detectors often rely on dataset-specific stylistic cues rather than stable signals of machine authorship. Further investigation with in-depth error analysis exposes a fundamental tension in linguistic-feature-based AI text detection: the features that are most discriminative on in-domain data are also the features most susceptible to domain shift, formatting variation, and text-length effects. We believe that this knowledge helps build AI detectors that are robust across different settings. To support replication and practical use, we release an open-source Python package that returns both predictions and instance-level explanations for individual texts.

cs.CL↗

arXiv2025

From Idea to Implementation: Evaluating the Influence of Large Language Models in Software Development -- An Opinion Paper

The introduction of transformer architecture was a turning point in Natural Language Processing (NLP). Models based on the transformer architecture such as Bidirectional Encoder Representations from Transformers (BERT) and Generative Pre-Trained Transformer (GPT) have gained widespread popularity in various applications such as software development and education. The availability of Large Language Models (LLMs) such as ChatGPT and Bard to the general public has showcased the tremendous potential of these models and encouraged their integration into various domains such as software development for tasks such as code generation, debugging, and documentation generation. In this study, opinions from 11 experts regarding their experience with LLMs for software development have been gathered and analysed to draw insights that can guide successful and responsible integration. The overall opinion of the experts is positive, with the experts identifying advantages such as increase in productivity and reduced coding time. Potential concerns and challenges such as risk of over-dependence and ethical considerations have also been highlighted.

cs.AI↗

arXiv2025

Exploring the Impact of Generative Artificial Intelligence in Education: A Thematic Analysis

The recent advancements in Generative Artificial intelligence (GenAI) technology have been transformative for the field of education. Large Language Models (LLMs) such as ChatGPT and Bard can be leveraged to automate boilerplate tasks, create content for personalised teaching, and handle repetitive tasks to allow more time for creative thinking. However, it is important to develop guidelines, policies, and assessment methods in the education sector to ensure the responsible integration of these tools. In this article, thematic analysis has been performed on seven essays obtained from professionals in the education sector to understand the advantages and pitfalls of using GenAI models such as ChatGPT and Bard in education. Exploratory Data Analysis (EDA) has been performed on the essays to extract further insights from the text. The study found several themes which highlight benefits and drawbacks of GenAI tools, as well as suggestions to overcome these limitations and ensure that students are using these tools in a responsible and ethical manner.

cs.AI↗

arXiv2024

Survey on Plagiarism Detection in Large Language Models: The Impact of ChatGPT and Gemini on Academic Integrity

The rise of Large Language Models (LLMs) such as ChatGPT and Gemini has posed new challenges for the academic community. With the help of these models, students can easily complete their assignments and exams, while educators struggle to detect AI-generated content. This has led to a surge in academic misconduct, as students present work generated by LLMs as their own, without putting in the effort required for learning. As AI tools become more advanced and produce increasingly human-like text, detecting such content becomes more challenging. This development has significantly impacted the academic world, where many educators are finding it difficult to adapt their assessment methods to this challenge. This research first demonstrates how LLMs have increased academic dishonesty, and then reviews state-of-the-art solutions for academic plagiarism in detail. A survey of datasets, algorithms, tools, and evasion strategies for plagiarism detection has been conducted, focusing on how LLMs and AI-generated content (AIGC) detection have affected this area. The survey aims to identify the gaps in existing solutions. Lastly, potential long-term solutions are presented to address the issue of academic plagiarism using LLMs based on AI tools and educational approaches in an ever-changing world.

cs.CY↗

arXiv2023

A Deep Learning Model for Heterogeneous Dataset Analysis -- Application to Winter Wheat Crop Yield Prediction

Western countries rely heavily on wheat, and yield prediction is crucial. Time-series deep learning models, such as Long Short Term Memory (LSTM), have already been explored and applied to yield prediction. Existing literature reported that they perform better than traditional Machine Learning (ML) models. However, the existing LSTM cannot handle heterogeneous datasets (a combination of data which varies and remains static with time). In this paper, we propose an efficient deep learning model that can deal with heterogeneous datasets. We developed the system architecture and applied it to the real-world dataset in the digital agriculture area. We showed that it outperforms the existing ML models.

cs.LG↗

arXiv2023

Winter Wheat Crop Yield Prediction on Multiple Heterogeneous Datasets using Machine Learning

Winter wheat is one of the most important crops in the United Kingdom, and crop yield prediction is essential for the nation's food security. Several studies have employed machine learning (ML) techniques to predict crop yield on a county or farm-based level. The main objective of this study is to predict winter wheat crop yield using ML models on multiple heterogeneous datasets, i.e., soil and weather on a zone-based level. Experimental results demonstrated their impact when used alone and in combination. In addition, we employ numerous ML algorithms to emphasize the significance of data quality in any machine-learning strategy.

cs.LG↗

arXiv2022

Enhancing Legal Argument Mining with Domain Pre-training and Neural Networks

The contextual word embedding model, BERT, has proved its ability on downstream tasks with limited quantities of annotated data. BERT and its variants help to reduce the burden of complex annotation work in many interdisciplinary research areas, for example, legal argument mining in digital humanities. Argument mining aims to develop text analysis tools that can automatically retrieve arguments and identify relationships between argumentation clauses. Since argumentation is one of the key aspects of case law, argument mining tools for legal texts are applicable to both academic and non-academic legal research. Domain-specific BERT variants (pre-trained with corpora from a particular background) have also achieved strong performance in many tasks. To our knowledge, previous machine learning studies of argument mining on judicial case law still heavily rely on statistical models. In this paper, we provide a broad study of both classic and contextual embedding models and their performance on practical case law from the European Court of Human Rights (ECHR). During our study, we also explore a number of neural networks when being combined with different embeddings. Our experiments provide a comprehensive overview of a variety of approaches to the legal argument mining task. We conclude that domain pre-trained transformer models have great potential in this area, although traditional embeddings can also achieve strong performance when combined with additional neural network layers.

cs.CL↗

arXiv2021

UCD-CS at TREC 2021 Incident Streams Track

In recent years, the task of mining important information from social media posts during crises has become a focus of research for the purposes of assisting emergency response (ES). The TREC Incident Streams (IS) track is a research challenge organised for this purpose. The track asks participating systems to both classify a stream of crisis-related tweets into humanitarian aid related information types and estimate their importance regarding criticality. The former refers to a multi-label information type classification task and the latter refers to a priority estimation task. In this paper, we report on the participation of the University College Dublin School of Computer Science (UCD-CS) in TREC-IS 2021. We explored a variety of approaches, including simple machine learning algorithms, multi-task learning techniques, text augmentation, and ensemble approaches. The official evaluation results indicate that our runs achieve the highest scores in many metrics. To aid reproducibility, our code is publicly available at https://github.com/wangcongcong123/crisis-mtl.

cs.CL↗

arXiv2021

Transformer-based Multi-task Learning for Disaster Tweet Categorisation

Social media has enabled people to circulate information in a timely fashion, thus motivating people to post messages seeking help during crisis situations. These messages can contribute to the situational awareness of emergency responders, who have a need for them to be categorised according to information types (i.e. the type of aid services the messages are requesting). We introduce a transformer-based multi-task learning (MTL) technique for classifying information types and estimating the priority of these messages. We evaluate the effectiveness of our approach with a variety of metrics by submitting runs to the TREC Incident Streams (IS) track: a research initiative specifically designed for disaster tweet classification and prioritisation. The results demonstrate that our approach achieves competitive performance in most metrics as compared to other participating runs. Subsequently, we find that an ensemble approach combining disparate transformer encoders within our approach helps to improve the overall effectiveness to a significant extent, achieving state-of-the-art performance in almost every metric. We make the code publicly available so that our work can be reproduced and used as a baseline for the community for future work in this domain.

cs.CL↗

arXiv2021

Crisis Domain Adaptation Using Sequence-to-sequence Transformers

User-generated content (UGC) on social media can act as a key source of information for emergency responders in crisis situations. However, due to the volume concerned, computational techniques are needed to effectively filter and prioritise this content as it arises during emerging events. In the literature, these techniques are trained using annotated content from previous crises. In this paper, we investigate how this prior knowledge can be best leveraged for new crises by examining the extent to which crisis events of a similar type are more suitable for adaptation to new events (cross-domain adaptation). Given the recent successes of transformers in various language processing tasks, we propose CAST: an approach for Crisis domain Adaptation leveraging Sequence-to-sequence Transformers. We evaluate CAST using two major crisis-related message classification datasets. Our experiments show that our CAST-based best run without using any target data achieves the state of the art performance in both in-domain and cross-domain contexts. Moreover, CAST is particularly effective in one-to-one cross-domain adaptation when trained with a larger language model. In many-to-one adaptation where multiple crises are jointly used as the source domain, CAST further improves its performance. In addition, we find that more similar events are more likely to bring better adaptation performance whereas fine-tuning using dissimilar events does not help for adaptation. To aid reproducibility, we open source our code to the community.

cs.CL↗

arXiv2021

Multi-task transfer learning for finding actionable information from crisis-related messages on social media

The Incident streams (IS) track is a research challenge aimed at finding important information from social media during crises for emergency response purposes. More specifically, given a stream of crisis-related tweets, the IS challenge asks a participating system to 1) classify what the types of users' concerns or needs are expressed in each tweet, known as the information type (IT) classification task and 2) estimate how critical each tweet is with regard to emergency response, known as the priority level prediction task. In this paper, we describe our multi-task transfer learning approach for this challenge. Our approach leverages state-of-the-art transformer models including both encoder-based models such as BERT and a sequence-to-sequence based T5 for joint transfer learning on the two tasks. Based on this approach, we submitted several runs to the track. The returned evaluation results show that our runs substantially outperform other participating runs in both IT classification and priority level prediction.

cs.CL↗

arXiv2020

Assessing the Influencing Factors on the Accuracy of Underage Facial Age Estimation

Swift response to the detection of endangered minors is an ongoing concern for law enforcement. Many child-focused investigations hinge on digital evidence discovery and analysis. Automated age estimation techniques are needed to aid in these investigations to expedite this evidence discovery process, and decrease investigator exposure to traumatic material. Automated techniques also show promise in decreasing the overflowing backlog of evidence obtained from increasing numbers of devices and online services. A lack of sufficient training data combined with natural human variance has been long hindering accurate automated age estimation -- especially for underage subjects. This paper presented a comprehensive evaluation of the performance of two cloud age estimation services (Amazon Web Service's Rekognition service and Microsoft Azure's Face API) against a dataset of over 21,800 underage subjects. The objective of this work is to evaluate the influence that certain human biometric factors, facial expressions, and image quality (i.e. blur, noise, exposure and resolution) have on the outcome of automated age estimation services. A thorough evaluation allows us to identify the most influential factors to be overcome in future age estimation systems.

cs.CV↗

arXiv2020

UCD-CS at W-NUT 2020 Shared Task-3: A Text to Text Approach for COVID-19 Event Extraction on Social Media

In this paper, we describe our approach in the shared task: COVID-19 event extraction from Twitter. The objective of this task is to extract answers from COVID-related tweets to a set of predefined slot-filling questions. Our approach treats the event extraction task as a question answering task by leveraging the transformer-based T5 text-to-text model. According to the official evaluation scores returned, namely F1, our submitted run achieves competitive performance compared to other participating runs (Top 3). However, we argue that this evaluation may underestimate the actual performance of runs based on text-generation. Although some such runs may answer the slot questions well, they may not be an exact string match for the gold standard answers. To measure the extent of this underestimation, we adopt a simple exact-answer transformation method aiming at converting the well-answered predictions to exactly-matched predictions. The results show that after this transformation our run overall reaches the same level of performance as the best participating run and state-of-the-art F1 scores in three of five COVID-related events. Our code is publicly available to aid reproducibility

cs.CL↗

arXiv2019

Improving Borderline Adulthood Facial Age Estimation through Ensemble Learning

Achieving high performance for facial age estimation with subjects in the borderline between adulthood and non-adulthood has always been a challenge. Several studies have used different approaches from the age of a baby to an elder adult and different datasets have been employed to measure the mean absolute error (MAE) ranging between 1.47 to 8 years. The weakness of the algorithms specifically in the borderline has been a motivation for this paper. In our approach, we have developed an ensemble technique that improves the accuracy of underage estimation in conjunction with our deep learning model (DS13K) that has been fine-tuned on the Deep Expectation (DEX) model. We have achieved an accuracy of 68% for the age group 16 to 17 years old, which is 4 times better than the DEX accuracy for such age range. We also present an evaluation of existing cloud-based and offline facial age prediction services, such as Amazon Rekognition, Microsoft Azure Cognitive Services, How-Old.net and DEX.

cs.CV↗

arXiv2018

Towards an Open Science Platform for the Evaluation of Data Fusion

Combining the results of different search engines in order to improve upon their performance has been the subject of many research papers. This has become known as the "Data Fusion" task, and has great promise in dealing with the vast quantity of unstructured textual data that is a feature of many Big Data scenarios. However, no universally-accepted evaluation methodology has emerged in the community. This makes it difficult to make meaningful comparisons between the various proposed techniques from reading the literature alone. Variations in the datasets, metrics, and baseline results have all contributed to this difficulty. This paper argues that a more unified approach is required, and that a centralised software platform should be developed to aid researchers in making comparisons between their algorithms and others. The desirable qualities of such a system have been identified and proposed, and an early prototype has been developed. Re-implementing algorithms published by other researchers is a great burden on those proposing new techniques. The prototype system has the potential to greatly reduce this burden and thus encourage more comparable results being generated and published more easily.

cs.IR↗

arXiv2017

Hierarchical Bloom Filter Trees for Approximate Matching

Bytewise approximate matching algorithms have in recent years shown significant promise in de- tecting files that are similar at the byte level. This is very useful for digital forensic investigators, who are regularly faced with the problem of searching through a seized device for pertinent data. A common scenario is where an investigator is in possession of a collection of "known-illegal" files (e.g. a collection of child abuse material) and wishes to find whether copies of these are stored on the seized device. Approximate matching addresses shortcomings in traditional hashing, which can only find identical files, by also being able to deal with cases of merged files, embedded files, partial files, or if a file has been changed in any way. Most approximate matching algorithms work by comparing pairs of files, which is not a scalable approach when faced with large corpora. This paper demonstrates the effectiveness of using a "Hierarchical Bloom Filter Tree" (HBFT) data structure to reduce the running time of collection-against-collection matching, with a specific focus on the MRSH-v2 algorithm. Three experiments are discussed, which explore the effects of different configurations of HBFTs. The proposed approach dramatically reduces the number of pairwise comparisons required, and demonstrates substantial speed gains, while maintaining effectiveness.

cs.CR↗

arXiv2017

EviPlant: An efficient digital forensic challenge creation, manipulation and distribution solution

Education and training in digital forensics requires a variety of suitable challenge corpora containing realistic features including regular wear-and-tear, background noise, and the actual digital traces to be discovered during investigation. Typically, the creation of these challenges requires overly arduous effort on the part of the educator to ensure their viability. Once created, the challenge image needs to be stored and distributed to a class for practical training. This storage and distribution step requires significant time and resources and may not even be possible in an online/distance learning scenario due to the data sizes involved. As part of this paper, we introduce a more capable methodology and system as an alternative to current approaches. EviPlant is a system designed for the efficient creation, manipulation, storage and distribution of challenges for digital forensics education and training. The system relies on the initial distribution of base disk images, i.e., images containing solely base operating systems. In order to create challenges for students, educators can boot the base system, emulate the desired activity and perform a "diffing" of resultant image and the base image. This diffing process extracts the modified artefacts and associated metadata and stores them in an "evidence package". Evidence packages can be created for different personae, different wear-and-tear, different emulated crimes, etc., and multiple evidence packages can be distributed to students and integrated into the base images. A number of additional applications in digital forensic challenge creation for tool testing and validation, proficiency testing, and malware analysis are also discussed as a result of using EviPlant.

cs.CR↗