Emotional AI and EdTech: serving the public good?

ABSTRACT

Education Technology (EdTech) companies are deploying emotional AI to quantify social and emotional learning. Focusing on facial coding emotional AI that uses computer vision and algorithms to see, recognise, categorise and learn about facial expressions of emotion, this paper evaluates nascent usage of these technologies in education. To do this, it assesses the nature of child rights, the history and modern usage of face-based emotional AI, methodology and efficacy, and what this paper sees as a clash of private and public interests. Concern is shown to be two-fold: first is on method, especially given scope for material effects on students; second are ethical and legal concerns. While proposing a list of considerations for any implementation for these technologies in the classroom, the paper concludes that significant risks exist in deployment of these technologies in the classroom.     

How do apes ape?

In the wake of telling critiques of the foundations on which earlier conclusions were based, the last 15 years have witnessed a renaissance in the study of social learning in apes. As a result, we are able to review 31 experimental studies from this period in which social learning in chimpanzees, gorillas, and orangutans has been investigated. The principal question framed at the beginning of this era, Do apes ape? has been answered in the affirmative, at least in certain conditions. The more interesting question now is, thus, How do apes ape? Answering this question has engendered richer taxonomies of the range of social-learning processes at work and new methodologies to uncover them. Together, these studies suggest that apes ape by employing a portfolio of alternative social-learning processes in flexibly adaptive ways, in conjunction with nonsocial learning. We conclude by sketching the kind of decision tree that appears to underlie the deployment of these alternatives.     

Video diary: a method for exploring learning dispositions

This paper presents an innovative method for qualitative research in education. The video dairy technique was developed and used alongside other qualitative methods to map the socio‐cultural landscape of primary‐school children's learning dispositions prior to their transfer to secondary school. The larger project comprised a case study of a group of children transferring from one Year 6 classroom to their local comprehensive secondary school, with particular reference to their learning of mathematics. Theoretically framed by Pierre Bourdieu's sociological tools, the study explored the impact of three fields—school, family and peer group—upon children's learning dispositions and trajectories at this school interface. The alternative perspectives afforded by the diary entries highlighted the inter‐relatedness of the three fields of interest, often in ways that had not been anticipated or considered. I will explain practical aspects of the process and consider some of the challenges that might be experienced by researchers using video diary techniques.     

Students' Understandings of Gravity in an Orbiting Space-Ship

We report on an investigation of students' ideas about gravity after a semester of instruction in physics at university. There are two aspects to the study which was concerned with students' answers to a carefully designed qualitative examination question on gravity. The first aspect is a classification of the answers and a comparative study of the ways the problem was tackled by two large groups of students who had different backgrounds in physics and were exposed to different teaching styles. The second aspect is to investigate how students link concepts to solve the problem. We used a phenomenographic analysis of student responses to extract patterns of reasoning and alternative conceptions behind the solutions. We found no differences between the classes of answers given by students in the two courses. Our analysis also identifies a hierarchy in the complexity of the hypothetical reasoning pathways, which we interpret as reflecting the ways in which students may link concepts and resolve conflicts as they solve the problem. The hypothetical reasoning pathways may help educators to develop instructional material or lecture room dialogue in order to tease out key issues. An unexpected finding is that there is a discrepancy between our conclusion that the two groups of answers are similar and the distribution of marks awarded by the examiner – which implies that the quality of the answers is different for the two groups.     

The Conconi test is not valid for estimation of the lactate turnpoint in runners

Conconi et al. (1982) reported that an observed deviation from linearity in the heart rate-running velocity relationship determined during a field test in runners coincided with the ‘lactate threshold’. The aim of this study was to assess the validity of the original Conconi test using conventional incremental and constant-load laboratory protocols. Fourteen trained male distance runners (mean ± s : age 22.6 ±3.4 years; body mass 67.6±4.8 kg; peak [Vdot] O 2 66.3 ± 4.7 ml kg -1 min -1) performed a standard multi-stage test for determination of lactate turnpoint and a Conconi test on a motorized treadmill. A deviation from linearity in heart rate was observed in nine subjects. Significant differences were found to exist between running velocity at the lactate turnpoint (4.39 ± 0.20 ms -1) and at deviation from linear heart rate (5.08 ± 0.25 ms -1) (P < 0.01), and between heart rate at the lactate turnpoint (172 ± 10 beats min -1) and at deviation from linearity (186 ± 9 beats min -1) (P < 0.01). When deviation of heart rate from linearity was evident, it occurred at a systematically higher intensity than the lactate turnpoint and at approximately 95% of maximum heart rate. These results were confirmed by the physiological responses of seven subjects, who performed two constant-velocity treadmill runs at 0.14 ms -1 below the running velocity at the lactate turnpoint and that at which the heart rate deviated from linearity. For the lactate turnpoint trial, the prescribed 30 min exercise period was completed by all runners (terminal blood lactate concentration of 2.4 ± 0.5 mM ), while the duration attained in the trial for which heart rate deviated from linearity was 15.9 ± 6.7 min (terminal blood lactate concentration of 8.1 ± 1.8 mM). We concluded that the Conconi test is invalid for the non-invasive determination of the lactate turnpoint and that the deviation of heart rate from linearity represents the start of the plateau at maximal heart rate, the expression of which is dependent upon the specifics of the Conconi test protocol.     

The Effects of Training and Position on Judges' Ratings of a Gymnastic Event

The purpose of this study was to determine the effects of training and position on judges' ratings of a gymnastic event. Twenty-four women judged the performance of the squat vault by 40 college women. The judges were divided into four groups. Two groups received written training, and two groups, extensive training. After completing their training, one written group and one extensive group viewed the performances from the side of the vaulting box; the other two groups viewed from the front of the vaulting box. Judge position and judge training were found to have significant effects upon judges' ratings. Significantly lower ratings were given by the judges who viewed the vault from the front and had received extensive training.     

Performance tolerance and boat set‐up in elite sprint Kayaking

The aim of this study was to examine the inter‐relationship between athlete morphology, equipment set‐up and performance in elite sprint kayaking. Correlations applied to data from the 2000 Olympics were used to select the most important links between morphology and boat set‐up — paddle grip width and foot‐bar distance. Associations between body size and the above selected equipment set‐ups were calculated using a Pearson correlation matrix, to facilitate the logical selection of independent variables as input for regression analyses. Significant (p < 0.01) regression equations were developed for the prediction of foot‐bar distance (r² = 0.589: standard error of estimate (SEE) = 4.48) and paddle grip width (r² = 0.541: SEE = 3.08). Three national‐standard sprint kayakers used their preferred set‐up together with modifications of their predicted set‐up, derived from Olympic data, to test performance tolerance in sprint kayaking. Mean coefficients of multiple determination over three trials for the three paddlers of 0.91, 0.91 and 0.92 for left paddle force, right paddle force, and paddle angle at water entry, respectively, were recorded when using their preferred set‐up. These data showed that the paddlers produce consistent patterns of motion. The intervention of altering the boat set‐up resulted in varying changes to boat speed. The mean preferred speed for the three paddlers of 4.47 m/s was reduced by 0.07 and 0.10 m/s when the above boat set‐up was modified to a predicted and ‘predicted plus one standard deviation’ respectively. These changes in boat speed were the result of alterations in the mechanics of paddling technique.     

Forthcoming events

Quantitative analysis of sports performance has been shown to produce information that coaches can use within the coaching process to enhance performance. Traditional methods for quantifying sport performances are limited in their capacity to describe the complex interactions of events that occur within a performance over time. In this paper, we outline a new approach to the analysis of time-based event records and real-time behaviour records on sport performance known as T-pattern detection. The relevant elements of the T-pattern detection process are explained and exemplar data from the analysis of 13 soccer matches are presented to highlight the potential of this form of analysis. The results from soccer suggest that it is possible to identify new profiles for both individuals and teams based on the analysis of temporal behavioural patterns detected within the performances.     

Language in Art and Art in Language

The National Curriculum is bringing a systematic attention to the place of language in the teaching and learning of Art, but may be suppressing some of the liveliness of language in art. Art teaches a specialist vocabulary with benefits beyond the art lesson, but there are dangers (and opportunities) in the use of words in art which have different meanings elsewhere. Art rooms have traditionally promoted a rich variety of language uses, but new pressures could lead to formulaic didactic lessons with too little pupil discussion. One language use in art, from which English teachers could learn, is discussion about the aesthetic qualities of artefacts which pupils make or are shown. English teaching too often treats poems as documentaries, but art can teach pupils to use words for looking and thinking about artefacts in their own terms. A danger here, in art as in English, is that introducing a canon can encourage talking about art in second-hand language which does not connect with pupils’ experiences. Language would drive a model of pedagogy in which experience and perception inform the formation of new concepts, then new concepts inform the search for new experience and perception, in an ascending spiral of aesthetic understanding which could also be both a pleasure and an education of the feelings.     

Consequences of "minimal" group affiliations in children

Three experiments (total N=140) tested the hypothesis that 5-year-old children's membership in randomly assigned "minimal" groups would be sufficient to induce intergroup bias. Children were randomly assigned to groups and engaged in tasks involving judgments of unfamiliar in-group or out-group children. Despite an absence of information regarding the relative status of groups or any competitive context, in-group preferences were observed on explicit and implicit measures of attitude and resource allocation (Experiment 1), behavioral attribution, and expectations of reciprocity, with preferences persisting when groups were not described via a noun label (Experiment 2). In addition, children systematically distorted incoming information by preferentially encoding positive information about in-group members (Experiment 3). Implications for the developmental origins of intergroup bias are discussed.