Artificial intelligence ethics requires a united approach from policymakers, AI companies, and individuals, in the development, deployment, and use of these technologies. However, sometimes discussions can become fragmented because of the different levels of governance or because of different values, stakeholders, and actors involved. Recently, these conflicts became very visible, with such examples as the dismissal of AI ethics researcher Dr. Timnit Gebru from Google and the resignation of whistle-blower Frances Haugen from Facebook. Underpinning each debacle was a conflict between (...) the organisation’s economic and business interests and the morals of their employees. This paper will examine tensions between the ethics of AI organisations and the values of their employees, by providing an exploration of the AI ethics literature in this area, and a qualitative analysis of three workshops with AI developers and practitioners. Common ethical and social tensions will be discussed, along with proposals on how to avoid or reduce these conflicts in practice. Altogether, we suggest the following steps to help reduce ethical issues within AI organisations: improved and diverse ethics education and training within businesses; internal and external ethics auditing; the establishment of AI ethics ombudsmen, AI ethics review committees and an AI ethics watchdog; as well as access to trustworthy AI ethics whistle-blower organisations. (shrink)

The ethics of artificial intelligence is a widely discussed topic. There are numerous initiatives that aim to develop the principles and guidance to ensure that the development, deployment and use of AI are ethically acceptable. What is generally unclear is how organisations that make use of AI understand and address these ethical issues in practice. While there is an abundance of conceptual work on AI ethics, empirical insights are rare and often anecdotal. This paper fills the gap in our current (...) understanding of how organisations deal with AI ethics by presenting empirical findings collected using a set of ten case studies and providing an account of the cross-case analysis. The paper reviews the discussion of ethical issues of AI as well as mitigation strategies that have been proposed in the literature. Using this background, the cross-case analysis categorises the organisational responses that were observed in practice. The discussion shows that organisations are highly aware of the AI ethics debate and keen to engage with ethical issues proactively. However, they make use of only a relatively small subsection of the mitigation strategies proposed in the literature. These insights are of importance to organisations deploying or using AI, to the academic AI ethics debate, but maybe most valuable to policymakers involved in the current debate about suitable policy developments to address the ethical issues raised by AI. (shrink)

Background Informed consent is a requirement for all research. It is not, however, clear how much information is sufficient to make an informed decision about participation in research. Information on an online questionnaire about childhood development was provided through an unfolding electronic participant sheet in three levels of information. Methods 552 participants, who completed the web-based survey, accessed and spent time reading the participant information sheet (PIS) between July 2008 and November 2009. The information behaviour of the participants was investigated. (...) The first level contained less information than might be found on a standard PIS, the second level corresponded to a standard PIS, and the third contained more information than on a standard PIS. The actual time spent on reading the information provided in three incremental levels and the participants' evaluation of the information were calculated. Results 77% of the participants chose to access the first level of information, whereas 12% accessed the first two levels, 6% accessed all three levels of information and 23% participated without accessing information. The most accessed levels of information were those that corresponded to the average reading times. Conclusion The brief information provided in the first level was sufficient for participants to make informed decisions, while a sizeable minority of the participants chose not to access any information at all. This study adds to the debate about how much information is required to make a decision about participation in research and the results may help inform the future development of information sheets by providing data on participants' actual needs when deciding about questionnaire surveys. (shrink)

Many decades ago Patrick Suppes argued rather convincingly that theoretical hypotheses are not confronted with the direct, raw results of an experiment, rather, they are typically compared with models of data. What exactly is a data model however? And how do the interactions of particles at the subatomic scale give rise to the huge volumes of data that are then moulded into a polished data model? The aim of this paper is to answer these questions by presenting a detailed case (...) study of the construction of data models at the LHCb for testing Lepton Flavour Universality in rare decays of B-mesons. The close examination of the scientific practice at the LHCb leads to the following four main conclusions: raw data in their pure form are practically useless for the comparison of experimental results with theory, and processed data are in some cases epistemically more reliable, real and simulated data are involved in the co-production of the final data model and cannot be easily distinguished, theory-ladenness emerges at three different levels depending on the scope and the purpose for which background theory guides the overall experimental process and the overall process of acquiring and analysing data in high energy physics is too complicated to be fully captured by a generic methodological description of the experimental practice. (shrink)

What are scientific models? Philosophers of science have been trying to answer this question during the last three decades by putting forward a number of different proposals. Some say that models are best understood as abstract Platonic objects or fictional entities akin to Sherlock Holmes, while others focus on their mathematical nature and see them as set theoretical structures. Although each account has its own strengths in offering various insights on the nature of models, several objections have been raised against (...) these views which still remain unanswered, making the debate on the ontology of models seem unresolvable. The primary aim of this paper is to show that a large part of these difficulties stems from an inappropriate reading of the main question on the ontology of models as a purely metaphysical question. Building on Carnap, it is argued that the question of the ontology of scientific models is either an internal theoretical question within an already accepted linguistic framework or an external practical question regarding the choice of the most appropriate form of language in order to describe and explain the practice of scientific modelling. The main implication of this view is that the question of the ontology of models becomes a means of probing other related questions regarding the overall practice of scientific modelling, such as questions on the capacity of models to provide knowledge and the relation of models with background theories. (shrink)

We investigate the applicability of Hegerfeldts arguments on Quantum nonlocality in Quantum Electrodynamics following the work of Prigogine, Pronko, Petrosky, Ordonez and Karpov. We demonstrate the appearance of nonlocal effects at the level of quantum states. We show, however that the expectation values of some observables spread causally. Therefore the measurement of the nonlocality is questionable. We investigate an approach to classical measurement and conclude that the classical measurement cannot detect the “acausal” effects of the non-locality.

Constantin Caratheodory offered the first systematic and contradiction free formulation of thermodynamics on the basis of his mathematical work on Pfaff forms. Moreover, his work on measure theory provided the basis for later improved formulations of thermodynamics and physics of continua where extensive variables are measures and intensive variables are densities. Caratheodory was the first to see that measure theory and not topology is the natural tool to understand the difficulties (ergodicity, approach to equilibrium, irreversibility) in the Foundations of Statistical (...) Physics. He gave a measure-theoretic proof of Poincaré's recurrence theorem in 1919. This work paved the way for Birkhoff to identify later ergodicity as metric transitivity and for Koopman and von Neumann to introduce spectral analysis of dynamical systems in Hilbert spaces. Mixing provided an explanation of the approach to equilibrium but not of irreversibility. The recent extension of spectral theory of dynamical systems to locally convex spaces, achieved by the Brussels–Austin groups, gives new nontrivial time asymmetric spectral decompositions for unstable and/or non-integrable systems. In this way irreversibility is resolved in a natural way. (shrink)

Aim The aim is to understand current research into the impact of undergraduate nursing education on the development of professional values. Background Values are evident in the professional standards for nurses and the guidelines and healthcare policies of many countries. These professional values guide decisions and behaviour and are recognised as an essential component in the professions ability to provide safe and professional care. This literature review presents the current research on the impact of education on professional values in undergraduate (...) nurse education. Design An integrative review of the findings was conducted to provide insight into the current research on the professional value development in undergraduate nurses. Data sources CINAHL, PubMed and Scopus. Review methods A literature search was undertaken within defined date parameters 2010–2021 using a systematic approach. The Preferred Reporting Items for Systematic Reviews and Meta-analysis guide PRISMA was used to guide and illustrate the process. Papers were assessed for quality using the Mixed Methods Appraisal Tool. Results Two distinct areas of inquiry were identified. (a) Changes in professional values as an outcome of undergraduate nursing education or (b) changes in professional values as an outcome of specifically designed educational content. These areas were further explored to better understand the influences of undergraduate education on students’ professional values. Conclusion There is a lack of evidence in the literature to support the premise that professional values develop in line with academic year progression; however, there is strong evidence to support the inclusion of explicit learning in undergraduate education that engages students in education specifically designed to explore and develop professional values. (shrink)

The logic of a physical system consists of the elementary observables of the system. We show that for chaotic systems the logic is not any more the classical Boolean lattice but a kind of fuzzy logic which we characterize for a class of chaotic maps. Among other interesting properties the fuzzy logic of chaos does not allow for infinite combinations of propositions. This fact reflects the instability of dynamics and it is shared also by quantum systems with diagonal singularity. We (...) also generalize the fuzzy implication to a probabilistic implication following the hint of von Neumann. In this way we can evaluate the probability of the validity of the logical inference. (shrink)

The work of the Brussels-Austin groups on irreversibility over the last years has shown that Quantum Large Poincaré systems with diagonal singularity lead to an extension of the conventional formulation of dynamics at the level of mixtures which is manifestly time asymmetric. States with diagonal singularity acquire meaning as linear fractionals over the involutive Banach algebra of operators with diagonal singularity. We show in this paper that the logic of quantum systems with diagonal singularity is not the conventional logic of (...) Hilbert space, because only finite combinations of prepositions are allowed. (shrink)

Current cosmological observations place little constraints on the nature of dark matter, allowing the development of a large number of models and various methods for probing their properties, which seem to provide ideal grounds for the employment of robustness arguments. In this article, the extent to which such arguments can be used to overcome various methodological and theoretical challenges is examined. The conclusion is that while robustness arguments have a limited scope in the context of dark matter research, they can (...) still be used for increasing the scientists’ confidence about the properties of specific models. (shrink)

The purpose of the present study was to test the hypothesis that goal orientation is associated with divergent forms of emotional reactivity under frustration. Goal orientations were assessed using bifurcations of performance goals described earlier. Physiological stress levels were measured via a blood volume pulse analysis after individuals were subjected to a computerized Stroop task using a malfunctioning mouse to induce enhanced frustration. The results indicated that performance-avoidance goals were associated with the highest levels of emotional reactivity, with normative outcome (...) goals being significantly more detrimental than ability goals. We concluded that the motivation to avoid failure or to outperform others is the most detrimental determinant of stress and needs to be avoided by all means. Instead, it is suggested that educators emphasize performance using personal best outcomes or by valuing engagement, deep processing and task completion. (shrink)

The plague of 541–542 ad started in Pelusium, a town in Egypt, and reached Constantinople in 542 ad. The following series summarizes the event and its repeated appearance until 746 ad.

The purpose of the present study was to test the measurement invariance of the baby pediatric symptom checklist across gender and age as a means to provide for valid comparisons in point estimates across groups. A secondary goal involved confirming the earlier identified factor structure and re-examining the presence of differentially item functioning in the BPSC across grouping variables. Participants were 601 children aged below 1 year and 1 year to 12 months. Data were collected as part of the National (...) Health Interview Survey in 2020. Results related to gender indicated satisfying all five levels of measurement and structural invariance. When testing measurement invariance across age groups, a few modifications were necessary. Specifically, partial metric invariance was satisfied after freeing slope estimates of items 1 and 2, partial scalar after freeing intercept terms of items 6 and 10, and partial residual invariance through freeing error terms of items 2 and 12. These non-invariant estimates, however, provide support for partial invariance and do not invalidate the comparison of latent means. The BPSC can be used for the valid assessment of social–emotional skills in infants up to the age of 18 months. (shrink)

In educational and clinical settings, few norm-referenced tests have been utilized until now usually focusing on a single or a few language subcomponents, along with very few language rating scales for parents and educators. The need for a comprehensive language assessment tool for preschool and early school years children which could form the basis for valid and reliable screening and diagnostic decisions, led to the development of a new norm-referenced digital tool called Logometro®. The aim of the present study is (...) to describe Logometro® as well as its psychometric characteristics. Logometro® evaluates an array of oral language skills across the different language domains such as phonological awareness, listening comprehension, vocabulary knowledge, narrative speech, morphological awareness, pragmatics, as well emergent literacy skills in Greek-speaking 4–7 years old children. More specifically, Logometro® has been designed in order to: map individual language development paths as well as difficulties, provide a descriptive profile of children’s oral language and emergent literacy skills, and assist in the identification of children who are at risk for Developmental Language Disorder or Specific Learning Disabilities. The sample consisted of 926 children aged from 4 to 7 years, which were recruited from diverse geographical provinces and represented a variety of socioeconomic backgrounds in Greece. Eight hundred participants were typically developing children, 126 children represented children with Special Educational Needs, and 126 children were typically developing peers matched for gender and age with the clinical groups. The administration lasted 90 min, depending on the participant’s age and competence. Validity as well as internal consistency and test–retest reliability were assessed. Results indicated that Logometro® is characterized by good psychometric properties and can constitute a norm-referenced battery of oral language and emergent literacy skills. It could be used to inform the professionals as well as the researchers about a child’s language strengths and weaknesses and form the basis on which they can design an appropriate individualized intervention if needed. (shrink)

Default logic is computationally expensive. One of the most promising ways of easing this problem and developing powerful implementations is to split a default theory into smaller parts and compute extensions in a modular, local way. This paper compares two recent approaches, Turner's splitting and Cholewinski's stratification. It shows that the approaches are closely related – in fact the former can be viewed as a special case of the latter.

IntroductionThe COVID-19 pandemic has led to unprecedented strain to healthcare systems worldwide and posed unique challenges to the healthcare professionals and the general public.ObjectivesThe aim of this study is to evaluate the impact of COVID-19 on the mental health, behavioral, and physical wellbeing of HCPs in the early and mid-term periods of the pandemic in comparison to non-HCPs. Thus, facilitating and guiding optimum planning and delivery of support to HCPs.Methods and AnalysisAn observational cross-sectional survey and cohort study aiming to enroll (...) over 1050 participants. Study questionnaires will be completed at baseline and after 6-weeks and 4-months. Recruitment initiated July 2020. The study was designed in London, United Kingdom, but open to participants worldwide. Baseline: Questionnaires comprising of validated self-administered screening tools for depression, anxiety, sleep-related issues, wellbeing, and burnout. The questionnaires also explore changes in behavior and physical wellbeing of the participants. In addition, associations of these mental health and behavioral factors with work-related factors and support will be explored. Six-weeks and 4-months follow-up: Follow-up questionnaires will assess change in symptoms of anxiety and depression, sleep disorders, use of alcohol and other substances, behavioral or interpersonal relationship changes. Physical wellbeing will be assessed through the presence of suspected or confirmed COVID-19 infection and absence from work. We will also evaluate the impact of variable provision of personal protection equipment, extended working hours, and concern for the wellbeing of family members, anxiety levels, and evidence of burnout.Statistical ConsiderationsThe study has 80% power to detect a 10% difference of combined depression and/or anxiety symptoms between the groups using two-sided type 1 error at 0.05 at baseline. Assuming that only 50% of these HCPs agree to be a part of a cohort survey, we will have 80% power to detect around 12% difference in the two groups in reported physical symptoms, or prevalence of depression and/or anxiety at the end of the study.EthicsThe study was approved by the Cambridge East, Research Ethics Committee.Trial Registration NumberClinicalTrials.gov, NCT04433260. (shrink)

This study investigates the ethical use of Big Data and Artificial Intelligence (AI) technologies (BD + AI)—using an empirical approach. The paper categorises the current literature and presents a multi-case study of 'on-the-ground' ethical issues that uses qualitative tools to analyse findings from ten targeted case-studies from a range of domains. The analysis coalesces identified singular ethical issues, (from the literature), into clusters to offer a comparison with the proposed classification in the literature. The results show that despite the variety (...) of different social domains, fields, and applications of AI, there is overlap and correlation between the organisations’ ethical concerns. This more detailed understanding of ethics in AI + BD is required to ensure that the multitude of suggested ways of addressing them can be targeted and succeed in mitigating the pertinent ethical issues that are often discussed in the literature. (shrink)

Open peer commentary on the article “Learning about Learning with Teachers and Young Children” by Chrystalla Papademetri-Kachrimani. Upshot: The activities that Papademetri-Kachrimani presents in her stories create situations that lead to unexpected results, thus opening the potential for learning about learning in teachers’ professional development. These integrate modeling-based learning - arguably a form of constructionism -, and allow learners to move back-and-forth between representations in order to develop strategies and rules.

Open peer commentary on the article “Learning about Learning with Teachers and Young Children” by Chrystalla Papademetri-Kachrimani. Upshot: In my commentary, I focus on the concept of surprise underlying the design of the learning experience presented in Papademetri-Kachrimani’s target article. I treat surprise as a concept that integrates the creative, open and non-predictable characteristics of constructionist teaching and learning. In my analysis, I show that current technological and societal developments have made these ideas of constructionism more relevant than ever. (...) Within this rendering, it becomes clear that there is a need to revisit the position of constructionism in education. (shrink)

Fixpoint semantics are provided for ambiguity blocking and propagating variants of Nute's defeasible logic. The semantics are based upon the well-founded semantics for logic programs. It is shown that the logics are sound with respect to their counterpart semantics and complete for locally finite theories. Unlike some other nonmonotonic reasoning formalisms such as Reiter's default logic, the two defeasible logics are directly skeptical and so reject floating conclusions. For defeasible theories with transitive priorities on defeasible rules, the logics are shown (...) to satisfy versions of Cut and Cautious Monotony. For theories with either conflict sets closed under strict rules or strict rules closed under transposition, a form of Consistency Preservation is shown to hold. The differences between the two logics and other variants of defeasible logic—specifically those presented by Billington, Antoniou, Governatori, and Maher—are discussed. (shrink)