The rapid irruption of AI is a double-edged sword. On the one hand, it poses a high existential risk to humanity, much higher than climate change. On the other hand, it provides an enormous opportunity to improve wellbeing. To handle this double-edged sword successfully, educators must prepare students for a future where citizens repeatedly interact with artificial agents (AAs). Unlike humans, AAs lack fear of death and social exclusion. They do not have family, attachment, or a sense of belonging. However, they have increasingly better abilities to read our minds. To counteract this negative side of AI, students need to understand the hidden psychology of AAs. They need to learn the core computational algorithms that drive the social behavior of AAs. To tackle this challenge, we have been developing creative coloring activities for elementary school students that teach them core AI algorithms. Contrary to multiplechoice questions, these are activities with millions of correct possible solutions. In addition, it is highly improbable that they will be solved by chance. Moreover, in these activities, students must explain their solutions in writing and pose similar problems to their peers. However, given the high complexity of these creative tasks, teachers need support to assess students' work. Thus, we developed an app to support teachers based on our experiences working with these coloring activities with teachers in underrepresented regions such as Chile and Peru and countries of the Southeast Asia Ministers of Education Organization (SEAMEO). The app helps them assess these rich multi-modal collaborative coloring activities by leveraging the power of large language models (LLMs). We present some key features of the app and discuss how it facilitates rapid and cost-effective implementation of these rich but demanding coloring activities.

Acoustic features and machine learning models have been recently proposed as promising tools to analyze lessons. Furthermore, acoustic patterns, both in the time and spectral domain, have been found to be related to teacher pedagogical practices. Nonetheless, most of previous work relies on expensive or third party equipment, limiting its scalability, and additionally, it is mainly used for diarization. Instead, in this work we present a cost-effective approach to identify teachers’ practices according to three categories (Presenting, Administration, and Guiding) which are compiled from the Classroom Observation Protocol for Undergraduate STEM. Particularly, we record teachers’ lessons using low-cost microphones connected to their smartphones. We then compute the mean and standard deviation of the amplitude, Mel spectrogram, and Mel Frequency Cepstral coefficients of the recordings to train supervised models for the task of predicting three categories compiled from the Classroom Observation Protocol for Undergraduate STEM. We found that spectral features perform better at the task of predicting teachers’ activities along the lessons and that our models can predict the presence of the two most common teaching practices with over 80% of accuracy and good discriminative power. Finally, with these models, we found that using audio obtained from the teachers’ smartphones it is also possible to automatically discriminate between sessions where students are using or not an online platform. This approach is important for teachers and other stakeholders who could use an automatic and cost-effective tool for analyzing teaching practices.

In this paper we explore the impact of an inter-classrooms math tournament implemented through internet. The strategy is to increase learning through intra-classroom collaboration generated by inter-classroom competition. Ten fourth grade classes with all their students from eight schools participated. During previous weeks students practiced on-line and played a cloud based board game designed to learn word problems. Afterwards, all students participated on an inter-classroom tournament. They played on-line synchronously during 60 min. The game was played in dyads formed from different schools. The list of each classroom average score was published every 5 min on each student computer. We found an important social facilitation effect: a significant improvement on the performance of male students weak on math, and therefore a reduction on the performance gap between mathematically weak and strong male students. The improvement of female students weak on math was also significant but lower.

Attracting students’ visual attention is critical in order for teachers to teach classes, communicate core concepts and emotionally connect with their students. In this paper we analyze two months of video recordings taken from a fourth grade class in a vulnerable school, where, every day, a sample of 3 students wore a mini video camera mounted on eyeglasses. We looked for scenes from the recordings where the teacher appears in the students’ visual field, and computed the average duration of each event. We found that the student’s gaze on the teacher lasted 44.9 % longer when the teacher gestured than when he did not, with an effect size (Cohen’s d) of 0.69. The data also reveals different effects for gender, subject matter, and student Grade Point Average (GPA). The effect of teacher gesturing on students with a low GPA is higher than on students’ with a high GPA. These findings may have broad significance for improving teaching practices.

Sustained improvement in student achievement on national standardized tests for low socio economic status (SES) districts is critical for reducing gaps in educational inequality. We report the results of 3 years of implementation of an ICT web-based learning environment in all 11 public schools of a low SES urban district in Chile. This includes 43 fourth grade classes and 1,355 students. This is a Computer Aided Instruction program that promotes whole class collaborative learning with peer support. Effect size on the national standardized fourth grade math test is 0.33, which is three times the national improvement level over the same period and five times the improvement made by a neighboring district with a similar population. On the other hand, the same students did not make any improvements on the national standardized language test. Since each class was taught by the same teacher, only without ICT, we can therefore discount the teacher effect.

We study two important features of the mechanisms living organisms seem to use to solve recurrent problems when able to choose strategies from a known set. The first one is forgetting and its influence on the ability of the organism to learn the chance of success of the known strategies. The other feature is selection of strategies according to their relative strengths. Specifically, we compare exponential and hyperbolic forgetting models, and we prove that when the agent has only one strategy available, the estimates for the strategy success rate using the exponential model never converge (in probabilistic terms), whereas the ones using the hyperbolic model converge almost surely. When more strategies are available and proportional selection is used, we prove several results that generalize the one strategy case.