MWMattias Wahde

Mattias Wahde

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mattias Wahde has a total of 31 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A note on the detection of LLM-generated political speeches

    by Minerva Suvanto, Mattias Wahde

    We use a recent data set (ParliaBench) containing political speeches as well as LLM-generated versions of such speeches to investigate whether the latter can be detected as being LLM-generated. We find that, for this data set, a linear, interpretable classifier can easily do so, achieving an F1 score of around 0.96 - 0.98. The ability of a classifier to distinguish LLM-generated text from human-written text is significant since it has many potential uses, for example in detecting potential undisclosed use of LLM-generated text in political speeches.

  • Open Access

    Interpretable Text Classification Applied to the Detection of LLM-generated Creative Writing

    by Minerva Suvanto, Andrea McGlinchey, Mattias Wahde, Peter J Barclay

    We consider the problem of distinguishing human-written creative fiction (excerpts from novels) from similar text generated by an LLM. Our results show that, while human observers perform poorly (near chance levels) on this binary classification task, a variety of machine-learning models achieve accuracy in the range 0.93 - 0.98 over a previously unseen test set, even using only short samples and single-token (unigram) features. We therefore employ an inherently interpretable (linear) classifier (with a test accuracy of 0.98), in order to elucidate the underlying reasons for this high accuracy. In our analysis, we identify specific unigram features indicative of LLM-generated text, one of the most important being that the LLM tends to use a larger variety of synonyms, thereby skewing the probability distributions in a manner that is easy to detect for a machine learning classifier, yet very difficult for a human observer. Four additional explanation categories were also identified, namely, temporal drift, Americanisms, foreign language usage, and colloquialisms. As identification of the AI-generated text depends on a constellation of such features, the classification appears robust, and therefore not easy to circumvent by malicious actors intent on misrepresenting AI-generated text as human work.

  • Part-of-Speech Taggers Make Errors on Unambiguous Sentences

    by Minerva Suvanto, Mattias Wahde, Marco L. Della Vedova

  • Open Access

    An interpretable method for automated classification of spoken transcripts and written text

    by Mattias Wahde, Marco L. Della Vedova, Marco Virgolin, Minerva Suvanto

    We investigate the differences between spoken language (in the form of radio show transcripts) and written language (Wikipedia articles) in the context of text classification. We present a novel, interpretable method for text classification, involving a linear classifier using a large set of $$n-$$ n - gram features, and apply it to a newly generated data set with sentences originating either from spoken transcripts or written text. Our classifier reaches an accuracy less than 0.02 below that of a commonly used classifier (DistilBERT) based on deep neural networks (DNNs). Moreover, our classifier has an integrated measure of confidence, for assessing the reliability of a given classification. An online tool is provided for demonstrating our classifier, particularly its interpretable nature, which is a crucial feature in classification tasks involving high-stakes decision-making. We also study the capability of DistilBERT to carry out fill-in-the-blank tasks in either spoken or written text, and find it to perform similarly in both cases. Our main conclusion is that, with careful improvements, the performance gap between classical methods and DNN-based methods may be reduced significantly, such that the choice of classification method comes down to the need (if any) for interpretability.


  • Open Access

    DAISY: An Implementation of Five Core Principles for Transparent and Accountable Conversational AI

    by Mattias Wahde, Marco Virgolin

    We present a detailed implementation of five core principles for transparent and acccountable conversational AI, namely interpretability, inherent capability to explain, independent data, interactive learning, and inquisitiveness. This implementation is a dialogue manager called DAISY that serves as the core part of a conversational agent. We show how DAISY-based agents are trained with human-machine interaction, a process that also involves suggestions for generalization from the agent itself. Moreover, these agents are capable to provide a concise and clear explanation of the actions required to reach a conclusion. Deep neural networks (DNNs) are currently the de facto standard in conversational AI. We therefore formulate a comparison between DAISY-based agents and two methods that use DNNs, on two popular data sets involving multi-domain task-oriented dialogue. Specifically, we provide quantitative results related to entity retrieval and qualitative results in terms of the type of errors that may occur. The results show that DAISY-based agents achieve superior precision at the price of lower recall, an outcome that might be preferable in task-oriented settings. Ultimately, and especially in view of their high degree of interpretability, DAISY-based agents are a fundamentally different alternative to the currently popular DNN-based methods.


  • Open Access

    The five Is: Key principles for interpretable and safe conversational AI

    by Mattias Wahde, Marco Virgolin

    In this position paper, we present five key principles, namely interpretability, inherent capability to explain, independent data, interactive learning, and inquisitiveness, for the development of conversational AI that, unlike the currently popular black box approaches, is transparent and accountable. At present, there is a growing concern with the use of black box statistical language models: While displaying impressive average performance, such systems are also prone to occasional spectacular failures, for which there is no clear remedy. In an effort to initiate a discussion on possible alternatives, we outline and exemplify how our five principles enable the development of conversational AI systems that are transparent and thus safer for use. We also present some of the challenges inherent in the implementation of those principles.


  • Open Access

    Lidar–Camera Semi-Supervised Learning for Semantic Segmentation

    by Luca Caltagirone, Mauro Bellone, Lennart Svensson, Mattias Wahde, Raivo Sell

    In this work, we investigated two issues: (1) How the fusion of lidar and camera data can improve semantic segmentation performance compared with the individual sensor modalities in a supervised learning context; and (2) How fusion can also be leveraged for semi-supervised learning in order to further improve performance and to adapt to new domains without requiring any additional labelled data. A comparative study was carried out by providing an experimental evaluation on networks trained in different setups using various scenarios from sunny days to rainy night scenes. The networks were tested for challenging, and less common, scenarios where cameras or lidars individually would not provide a reliable prediction. Our results suggest that semi-supervised learning and fusion techniques increase the overall performance of the network in challenging scenarios using less data annotations.


  • Open Access

    Model Learning with Personalized Interpretability Estimation (ML-PIE)

    by Marco Virgolin, Andrea De Lorenzo, Francesca Randone, Eric Medvet, Mattias Wahde

    High-stakes applications require AI-generated models to be interpretable. Current algorithms for the synthesis of potentially interpretable models rely on objectives or regularization terms that represent interpretability only coarsely (e.g., model size) and are not designed for a specific user. Yet, interpretability is intrinsically subjective. In this paper, we propose an approach for the synthesis of models that are tailored to the user by enabling the user to steer the model synthesis process according to her or his preferences. We use a bi-objective evolutionary algorithm to synthesize models with trade-offs between accuracy and a user-specific notion of interpretability. The latter is estimated by a neural network that is trained concurrently to the evolution using the feedback of the user, which is collected using uncertainty-based active learning. To maximize usability, the user is only asked to tell, given two models at the time, which one is less complex. With experiments on two real-world datasets involving 61 participants, we find that our approach is capable of learning estimations of interpretability that can be very different for different users. Moreover, the users tend to prefer models found using the proposed approach over models found using non-personalized interpretability indices.

  • A Dialogue manager for task-oriented agents based on dialogue building-blocks and generic cognitive processing

    by Mattias Wahde

    This paper introduces a novel dialogue manager, called DAISY, for intelligent virtual agents. The proposed approach is based on two central concepts: (i) dialogue building-blocks, which offer a systematic approach to the representation and implementation of human-agent dialogue, and (ii) cognitive processing in the form of sequences of simple, generic cognitive actions. The generic nature of the cognitive actions makes it possible to represent a large variety of cognitive processing (e.g, retrieving and manipulating memory content) by using a rather small set of such actions. DAISY is illustrated by means of a specific example, namely an agent acting as an information system for travel and tourist information. The example highlights the usefulness of the systematic approach offered by DAISY.