9607 research outputs found
Sort by
The Uniqueness Trap
No full textProject managers and planners are highly prone to believing that their projects are one of a kind—partly because those that seem new and distinctive are more likely to win support. But research on 1,300-plus projects reveals that few, if any, actually are unique. The problem is, the perception of uniqueness causes managers to think there’s nothing to learn from other projects, which leads them to underestimate risk, make poor decisions, and blow through budgets and schedules. The cure is to always assume that someone, somewhere has undertaken a project like yours. If you can’t find an analogue, break your project into components, which may prove comparable with other projects. Then use forecasting and other risk assessment tools to avoid biases that undermine good choices
Modyn: Data-Centric Machine Learning Pipeline Orchestration
No full textIn real-world machine learning (ML) pipelines, datasets are continuously growing. Models must incorporate this new training data to improve generalization and adapt to potential distribution shifts. The cost of model retraining is proportional to how frequently the model is retrained and how much data it is trained on, which makes the naive approach of retraining from scratch each time impractical. We present Modyn, a data-centric end-to-end machine learning platform. Modyn's ML pipeline abstraction enables users to declaratively describe policies for continuously training a model on a growing dataset. Modyn pipelines allow users to apply data selection policies (to reduce the number of data points) and triggering policies (to reduce the number of trainings). Modyn executes and orchestrates these continuous ML training pipelines. The system is open-source and comes with an ecosystem of benchmark datasets, models, and tooling. We formally discuss how to measure the performance of ML pipelines by introducing the concept of composite models, enabling fair comparison of pipelines with different data selection and triggering policies. We empirically analyze how various data selection and triggering policies impact model accuracy, and also show that Modyn enables high throughput training with sample-level data selection
Homomorphism Tensors and Linear Equations
No full textLovász (1967) showed that two graphs and are isomorphic if and only ifthey are homomorphism indistinguishable over the class of all graphs, i.e. for every graph , the number of homomorphisms from to equals the number of homomorphisms from to . Recently, homomorphism indistinguishability over restricted classes of graphs such as bounded treewidth, bounded treedepth and planar graphs, has emerged as a surprisingly powerful framework for capturing diverse equivalence relations on graphs arising from logical equivalence and algebraic equation systems.In this paper, we provide a unified algebraic framework for such results by examining the linear-algebraic structure of tensors counting homomorphisms from labelled graphs. The existence of certain linear transformations between such homomorphism tensor subspaces can be interpreted both as homomorphism indistinguishability over a graph class and as feasibility of an equational system. Following this framework, we obtain characterisations of homomorphism indistinguishability over several natural graph classes, namely trees of bounded degree, graphs of bounded pathwidth (answering a question of Dell et al. (2018)), and graphs of bounded treedepth
Research Community Perspectives on “Intelligence” and Large Language Models
No full textDespite the widespread use of “artificial intelligence” (AI) framing in Natural Language Processing (NLP) research, it is not clear what researchers mean by “intelligence”. To that end, we present the results of a survey on the notion of “intelligence” among researchers and its role in the research agenda. The survey elicited complete responses from 303 researchers from a variety of fields including NLP, Machine Learning (ML), Cognitive Science, Linguistics, and Neuroscience. We identify 3 criteria of intelligence that the community agrees on the most: generalization, adaptability, & reasoning. Our results suggests that the perception of the current NLP systems as “intelligent” is a minority position (29%). Furthermore, only 16.2% of the respondents see developing intelligent systems as a research goal, and these respondents are more likely to consider the current systems intelligent
Essays on Validity of Human-Computer Interaction Experiments
Get PDFHuman–computer interaction (HCI) experiments have been used to understand human– computer interaction and to inform the design of interactive products. However, to be able to derive causal inferences and make generalizations from experimental outcomes, close attention must be paid to how experiments are designed. Small flaws in experimental design can pose serious threats to the validity of experimental studies. Yet only limited attention has been paid to validity in HCI experiments. Many disciplines have taken steps to understand and address validity problems, and there is a common view that HCI should simply follow these steps. In this thesis, I argue for the necessity of closer attention to experimental design from within HCI. Drawing on empirical evidence, I illustrate the importance of replication as a source of validation and cumulative knowledge in the field. I further argue that HCI sometimes fails to fully engage with developments from other disciplines. In particular, I demonstrate that some experimental artefacts well known in psychology have been overlooked, despite their relevance to HCI. I also argue that practices imported from psychology require careful adaptation to the HCI context, as, for example, control conditions commonly used in psychological experiments do not always translate well to computerized settings. Finally, I show how poor reporting of study analyses may call into question the credibility of research findings. Overall, this thesis aims to identify validity issues in human–computer experiments and provide recommendations for how future HCI experiments can mitigate validity threats.<br/
Urban Mobility
No full textIn this chapter, we discuss urban mobility from a complexity science perspective. First, we give an overview of the datasets that enable this approach, such as mobile phone records, location-based social network traces, or GPS trajectories from sensors installed on vehicles. We then review the empirical and theoretical understanding of the properties of human movements, including the distribution of travel distances and times, the entropy of trajectories, and the interplay between exploration and exploitation of locations. Next, we explain generative and predictive models of individual mobility, and their limitations due to intrinsic limits of predictability. Finally, we discuss urban transport from a systemic perspective, including system-wide challenges like ridesharing, multimodality, and sustainable transport
AmendMe: A Tabletop Game to Teach the European Parliament's Legislative Process
No full textThe European Parliament Role-Play Game (EP RPG) is a multiplayer simulation in which the players take on the role of Members of the European Parliament (MEP). The game aims to help young European citizens understand the EU law-making process and to increase their interest in politics. It exists in a physical version, available to play on-site in Brussels and in many other capital cities all over Europe, as well as a virtual version that can be played in any suitably sized location, provided there are at least 16 players and a stable Internet connection. While the existing versions of the EP RPG strive to reach different audiences, they may still not be equally accessible to all players. For example, playing the physical game requires advance planning and travel, which may be a challenge for schools in rural and remote areas. Meanwhile, the virtual game may be difficult to play at schools with an unstable internet connection or limited hardware resources, or by smaller groups who would like to play outside of class. This paper focuses on developing a tabletop version of the EP RPG optimised for accessibility to a broader range of players. The game design focuses on game elements that support discussion among players and has them reenact key aspects of the EU legislative process, such as decision-making, compromise, and voting. Feedback from the pilot playtesting was encouraging and confirmed that the game was engaging and educational, but revealed that its content and mechanics require further adjustment
Extracting Participation in Collective Action from Social Media
Get PDFSocial media play a key role in mobilizing collective action, holding the potential for studying the pathways that lead individuals to actively engage in addressing global challenges. However, quantitative research in this area has been limited by the absence of granular and large-scale ground truth about the level of participation in collective action among individual social media users. To address this limitation, we present a novel suite of text classifiers designed to identify expressions of participation in collective action from social media posts, in a topic-agnostic fashion. Grounded in the theoretical framework of social movement mobilization, our classification captures participation and categorizes it into four levels: recognizing collective issues, engaging in calls-to-action, expressing intention of action, and reporting active involvement. We constructed a labeled training dataset of Reddit comments through crowdsourcing, which we used to train BERT classifiers and fine-tune Llama3 models. Our findings show that smaller language models can reliably detect expressions of participation (weighted F1=0.71), and rival larger models in capturing nuanced levels of participation. By applying our methodology to Reddit, we illustrate its effectiveness as a robust tool for characterizing online communities in innovative ways compared to topic modeling, stance detection, and keyword-based methods. Our framework contributes to Computational Social Science research by providing a new source of reliable annotations useful for investigating the social dynamics of collective action
Can You Link Up With Treewidth?
No full textA central result by Marx [ToC '10] constructs k-vertex graphs H of maximum degree 3 such that n^o(k/log k) time algorithms for detecting colorful H-subgraphs would refute the Exponential-Time Hypothesis (ETH). This result is widely used to obtain almost-tight conditional lower bounds for parameterized problems under ETH.Our first contribution is a new and fully self-contained proof of this result that further simplifies a recent work by Karthik et al. [SOSA 2024]. In our proof, we introduce a novel graph parameter of independent interest, the linkage capacity γ(H), and show that detecting colorful H-subgraphs in time n^o(γ(H)) refutes ETH. Then, we use a simple construction of communication networks credited to Beneš to obtain k-vertex graphs of maximum degree 3 and linkage capacity Ω(k/log k), avoiding arguments involving expander graphs, which were required in previous papers. We also show that every graph H of treewidth t has linkage capacity Ω(t/log t), thus recovering a stronger result shown by Marx [ToC '10] with a simplified proof.Additionally, we obtain new tight lower bounds on the complexity of subgraph detection for certain types of patterns by analyzing their linkage capacity: We prove that almost all k-vertex graphs of polynomial average degree Ω(k^β) for β > 0 have linkage capacity Θ(k), which implies tight lower bounds for finding such patterns H. As an application of these results, we also obtain tight lower bounds for counting small induced subgraphs having a fixed property Φ, improving bounds from, e.g., [Roth et al., FOCS 2020]