Considerations for a More Ethical Approach to Data in AI: On Data Representation and Infrastructure

Baird, Alice; Schuller, Björn W.

doi:10.3389/fdata.2020.00025

Cited by 22 publications

(12 citation statements)

References 94 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nevertheless, pretraining such models from scratch can be time-consuming and computationally expensive, rendering it impractical for most users. Moreover, implementing this approach in educational contexts poses signi cant challenges, including potential issues with ne-tuning instabilities (Dodge et al, 2020; Lee, Cho, and Kang, 2019), and raises ethical concerns regarding diversity and representation (Baird and Schuller, 2020;Yan et al, 2023) given the inherent constraints of small and skewed datasets.…”

Section: Contextualized Word Embeddingmentioning

confidence: 99%

Extending a Pretrained Language Model (BERT) using an Ontological Perspective to Classify Students’ Scientific Expertise Level from Written Responses

Wang,

Haudek,

Manzanares

et al. 2024

Preprint

View full text Add to dashboard Cite

The complex and interdisciplinary nature of scientific concepts presents formidable challenges for students in developing their knowledge-in-use skills. The utilization of computerized analysis for evaluating students’ contextualized constructed responses offers a potential avenue for educators to develop personalized and scalable interventions, thus supporting the teaching and learning of science consistent with contemporary calls. While prior research in artificial intelligence has demonstrated the effectiveness of algorithms, including Bidirectional Encoder Representations from Transformers (BERT), in tasks like automated classifications of constructed responses, these efforts have predominantly leaned towards text-level features, often overlooking the exploration of conceptual ideas embedded in students’ responses from a cognitive perspective. Despite BERT’s performance in downstream tasks, challenges may arise in domain-specific tasks, particularly in establishing knowledge connections between specialized and open domains. These challenges become pronounced in small-scale and imbalanced educational datasets, where the available information for fine-tuning is frequently inadequate to capture task-specific nuances and contextual details. The primary objective of the present study is to investigate the effectiveness of a pretrained language model (BERT), when integrated with an ontological framework aligned with a contextualized science assessment, in classifying students’ expertise levels in scientific explanation. Our findings indicate that while pretrained language models such as BERT contribute to enhanced performance in language-related tasks within educational contexts, the incorporation of identifying domain-specific terms and extracting and substituting with their associated sibling terms in sentences through ontology-based systems can significantly improve classification model performance. Further, we qualitatively examined student responses and found that, as expected, the ontology framework identified and substituted key domain specific terms in student responses that led to more accurate predictive scores. The study explores the practical implementation of ontology in assessment evaluation to facilitate formative assessment and formulate instructional strategies.

show abstract

Section: Contextualized Word Embeddingmentioning

confidence: 99%

Extending a Pretrained Language Model (BERT) using an Ontological Perspective to Classify Students’ Scientific Expertise Level from Written Responses

Wang,

Haudek,

Manzanares

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Besides these entities, there are other initiatives aimed at ensuring quality assurance, such as Watchdogs (Ada Lovelace Institute, AI Now, and Open Government Partnership, 2021), ethical committees within technology companies (Candelon et al , 2022), decentralised data access (Baird and Schuller, 2020), algorithmic social contracts (Rahwan, 2018) and human rights impact assessments (Moss et al , 2021). However, these initiatives are less known to the public and do not communicate the company’s efforts towards FATE AI to the consumer.…”

Section: Potential Certifying Entities (Real World)mentioning

confidence: 99%

A consumer perspective of AI certification – the current certification landscape, consumer approval and directions for future research

Blösser,

Weihrauch

2023

EJM

View full text Add to dashboard Cite

Purpose In spite of the merits of artificial intelligence (AI) in marketing and social media, harm to consumers has prompted calls for AI auditing/certification. Understanding consumers’ approval of AI certification entities is vital for its effectiveness and companies’ choice of certification. This study aims to generate important insights into the consumer perspective of AI certifications and stimulate future research. Design/methodology/approach A literature and status-quo-driven search of the AI certification landscape identifies entities and related concepts. This study empirically explores consumer approval of the most discussed entities in four AI decision domains using an online experiment and outline a research agenda for AI certification in marketing/social media. Findings Trust in AI certification is complex. The empirical findings show that consumers seem to approve more of non-profit entities than for-profit entities, with the government approving the most. Research limitations/implications The introduction of AI certification to marketing/social media contributes to work on consumer trust and AI acceptance and structures AI certification research from outside marketing to facilitate future research on AI certification for marketing/social media scholars. Practical implications For businesses, the authors provide a first insight into consumer preferences for AI-certifying entities, guiding the choice of which entity to use. For policymakers, this work guides their ongoing discussion on “who should certify AI” from a consumer perspective. Originality/value To the best of the authors’ knowledge, this work is the first to introduce the topic of AI certification to the marketing/social media literature, provide a novel guideline to scholars and offer the first set of empirical studies examining consumer approval of AI certifications.

show abstract

“…The movements are recorded over the entire duration of the video sequence and sampled with a bin size of 0.25 Hz on an axis magnitude between -1 000 and 1 000. Every annotation was checked by an auditor using quantitative and qualitative measures to ensure a high quality (Baird and Schuller, 2020). The time required for annotation alone stands for more than 600 working hours (40 h video * 3 dimensions * 5 annotators per dimension).…”

Section: Emotion and Trustworthiness Signalsmentioning

confidence: 99%

An Estimation of Online Video User Engagement From Features of Time- and Value-Continuous, Dimensional Emotions

Stappen¹,

Baird²,

Lienhart³

et al. 2022

Front. Comput. Sci.

Self Cite

View full text Add to dashboard Cite

Portraying emotion and trustworthiness is known to increase the appeal of video content. However, the causal relationship between these signals and online user engagement is not well understood. This limited understanding is partly due to a scarcity in emotionally annotated data and the varied modalities which express user engagement online. In this contribution, we utilize a large dataset of YouTube review videos which includes ca. 600 h of dimensional arousal, valence and trustworthiness annotations. We investigate features extracted from these signals against various user engagement indicators including views, like/dislike ratio, as well as the sentiment of comments. In doing so, we identify the positive and negative influences which single features have, as well as interpretable patterns in each dimension which relate to user engagement. Our results demonstrate that smaller boundary ranges and fluctuations for arousal lead to an increase in user engagement. Furthermore, the extracted time-series features reveal significant (p < 0.05) correlations for each dimension, such as, count below signal mean (arousal), number of peaks (valence), and absolute energy (trustworthiness). From this, an effective combination of features is outlined for approaches aiming to automatically predict several user engagement indicators. In a user engagement prediction paradigm we compare all features against semi-automatic (cross-task), and automatic (task-specific) feature selection methods. These selected feature sets appear to outperform the usage of all features, e.g., using all features achieves 1.55 likes per day (Lp/d) mean absolute error from valence; this improves through semi-automatic and automatic selection to 1.33 and 1.23 Lp/d, respectively (data mean 9.72 Lp/d with a std. 28.75 Lp/d).

show abstract

Considerations for a More Ethical Approach to Data in AI: On Data Representation and Infrastructure

Cited by 22 publications

References 94 publications

Extending a Pretrained Language Model (BERT) using an Ontological Perspective to Classify Students’ Scientific Expertise Level from Written Responses

Extending a Pretrained Language Model (BERT) using an Ontological Perspective to Classify Students’ Scientific Expertise Level from Written Responses

A consumer perspective of AI certification – the current certification landscape, consumer approval and directions for future research

An Estimation of Online Video User Engagement From Features of Time- and Value-Continuous, Dimensional Emotions

Contact Info

Product

Resources

About