Growing research in sign language recognition, generation, and translation AI has been accompanied by calls for ethical development of such technologies. While these works are crucial to helping individual researchers do better, there is a notable lack of discussion of systemic biases or analysis of rhetoric that shape the research questions and methods in the field, especially as it remains dominated by hearing non-signing researchers. Therefore, we conduct a systematic review of 101 recent papers in sign language AI. Our analysis identifies significant biases in the current state of sign language AI research, including an overfocus on addressing perceived communication barriers, a lack of use of representative datasets, use of annotations lacking linguistic foundations, and development of methods that build on flawed models. We take the position that the field lacks meaningful input from Deaf stakeholders, and is instead driven by what decisions are the most convenient or perceived as important to hearing researchers. We end with a call to action: the field must make space for Deaf researchers to lead the conversation in sign language AI.
LINK
In this paper, we report on the initial results of an explorative study that aims to investigate the occurrence of cognitive biases when designers use generative AI in the ideation phase of a creative design process. When observing current AI models utilised as creative design tools, potential negative impacts on creativity can be identified, namely deepening already existing cognitive biases but also introducing new ones that might not have been present before. Within our study, we analysed the emergence of several cognitive biases and the possible appearance of a negative synergy when designers use generative AI tools in a creative ideation process. Additionally, we identified a new potential bias that emerges from interacting with AI tools, namely prompt bias.
DOCUMENT
In dit artikel worden de belangrijkste oorzaken beschreven van problemen bij het bepalen van en het omgaan met risico’s in het veiligheidsdomein: beperkte gecijferdheid en structurele denkfouten (biases). Aan de hand van een procesmodel wordt inzichtelijk gemaakt waar in het analyse- en besluitvormingsproces de schoen wringt. Het model brengt daarbij bestaande modellen van risicomanagement terug tot de twee fasen die voor het analyseren van problema-tische risicobeslissingen cruciaal zijn: de fase van risicoanalyse en die van risicobeslissing. Waar in de subdomeinen fysieke en sociale veiligheid doorgaans geheel eigen analysemodellen en -methoden gangbaar zijn, biedt het hier gepresenteerde procesmodel een generiek toepasbare benadering van problemen met risicobesluitvorming in het domein van de integrale veiligheid als geheel. Het overzicht van knelpunten is opgesteld aan de hand van de onderzoeksliteratuur en wordt geïllustreerd met behulp van praktijkvoorbeelden. Aan het eind van het artikel wordt stilgestaan bij de vraag of er voor de gesignaleerde problemen een remedie kan worden ont-wikkeld.
MULTIFILE
Als lector duurzame paardenhouderij en paardensport is Inga Wolframm verbonden aan Hogeschool van Hall Larenstein. In die hoedanigheid deed ze onderzoek naar de ‘biases’ bij de jurysport die het onderdeel dressuur met zich meebrengt.
LINK
This paper identifies some common and specific pitfalls in the development of sign language technologies targeted at deaf communities, with a specific focus on signing avatars. It makes the call to urgently interrogate some of the ideologies behind those technologies, including issues of ethical and responsible development. The paper addresses four separate and interlinked issues: ideologies about deaf people and mediated communication, bias in data sets and learning, user feedback, and applications of the technologies. The paper ends with several take away points for both technology developers and deaf NGOs. Technology developers should give more consideration to diversifying their team and working interdisciplinary, and be mindful of the biases that inevitably creep into data sets. There should also be a consideration of the technologies’ end users. Sign language interpreters are not the end users nor should they be seen as the benchmark for language use. Technology developers and deaf NGOs can engage in a dialogue about how to prioritize application domains and prioritize within application domains. Finally, deaf NGOs policy statements will need to take a longer view, and use avatars to think of a significantly better system compared to what sign language interpreting services can provide.
LINK
In the social sciences, especially in social psychology and sociology, fear of the unknown and the tendency to favour the group one belongs to are recognised phenomena. This chapter will look at elements of these scientific theories and establish a link between these elements and the initial findings of a study of Syrian refugees on the Dutch labour market. First, this study is described, after which a glimpse into the world of social psychology is provided. This chapter concludes by taking a wide perspective on Dutch society and by discussing some lessons that can be learned for a diverse future. Research and education can support society in highlighting the importance of diversity and in training people to recognise the added value of diversity. Everyone will need to put considerable effort into ensuring people are judged for their personal qualities. Lecturers and professionals will need to examine their own judgements and biases first, before they can help to shape the perceptions of students. Together with students, lecturers can promote the benefits of the contact hypothesis and the concept of interactional diversity.
DOCUMENT
The HCR-20V3 is a violence risk assessment tool that is widely used in forensic clinical practice for risk management planning. The predictive value of the tool, when used in court for legal decisionmaking, is not yet intensively been studied and questions about legal admissibility may arise. This article aims to provide legal and mental health practitioners with an overview of the strengths and weaknesses of the HCR-20V3 when applied in legal settings. The HCR-20V3 is described and discussed with respect to its psychometric properties for different groups and settings. Issues involving legal admissibility and potential biases when conducting violence risk assessments with the HCR-20V3 are outlined. To explore legal admissibility challenges with respect to the HCR-20V3, we searched case law databases since 2013 from Australia, Canada, Ireland, the Netherlands, New Zealand, the UK, and the USA. In total, we found 546 cases referring to the HCR-20/HCR-20V3. In these cases, the tool was rarely challenged (4.03%), and when challenged, it never resulted in a court decision that the risk assessment was inadmissible. Finally, we provide recommendations for legal practitioners for the cross-examination of risk assessments and recommendations for mental health professionals who conduct risk assessments and report to the court. We conclude with suggestions for future research with the HCR-20V3 to strengthen the evidence base for use of the instrument in legal contexts.
DOCUMENT
The study evaluated two speech recognition systems, Wav2vec2 and Whisper, for potential biases for Dutch speakers.Results obtained by evaluating on the JASMIN corpus revealed biases against non-native speakers, children, and the elderly,with (slightly) better performance for women. The study emphasizes the need for ASR systems to handle variations in speakingin order to reach equal performance among all users.
DOCUMENT
Harvard Business Review auteurs Bazerman, Loewenstein en Moore stellen in hun artikel “Why good accountants do bad audits” (2002) dat het werkelijke probleem bij accountantscontroles niet zit in bewuste fraude door accountants maar schuilt in onbewuste beoordelingsfouten. Zij beroepen zich enerzijds op literatuur die constateert dat professionals onbewuste beoordelingsfouten begaan en anderzijds beschrijven zij omgevingsfactoren in de accountancypraktijk die onbewuste beoordelingsfouten door accountants zeer waarschijnlijk maken. In beperkte mate constateren zij door middel van experimenten dat accountants onbewuste beoordelingsfouten begaan. Onze studie beoogt op dit pad verder te gaan. Een groep studenten wordt aan een experiment in een accountancy-achtige omgeving onderworpen. Zij krijgen een waarderingsprobleem voorgelegd. De studenten blijken gevoelig voor suggesties in de hoogte van de waardering van het te waarderen object. Ook zijn ze gevoelig voor de presentatie van het object in de vorm van een bezitting of een schuld. De eerste onbewuste beoordelingsfout staat bekend als het anchoring effect; de tweede als het framing effect. We concluderen dat beide onbewuste beoordelingsfouten het onafhankelijke en objectieve oordeel van de accountant kunnen bedreigen.
DOCUMENT
Bij het ontwerpen en inrichten van ICT-architectuur worden beslissingen genomen die als vanzelfsprekend worden gezien, maar die wel degelijk een keuze inhouden en ethisch relevant zijn. Anke van Gorp pleit ervoor dat ICT-architecten deze keuzes ter discussie durven te stellen.
LINK
