An Interaction Design for Machine Teaching to Develop AI Tutors

Weitekamp, D. P.; Harpstead, Erik; Koedinger, Ken

doi:10.1145/3313831.3376226

Cited by 42 publications

(13 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We believe that learning technology should be humancentered because it aims at teaching and interactive activities. The HCAI approach now taking shape aims to enhance human capabilities, such as by allowing teachers to build their own computerized lessons using insights gathered from an AI tutoring system (Weitekamp et al, 2020). AI-supported learning environments therefore must not only focus on performance, but also human emotions and outcomes should be main concerns.…”

Section: A Human-centered Ai Approach In Educationmentioning

confidence: 99%

Reinvigorating the Discourse on Human-Centered Artificial Intelligence in Educational Technologies

Renz¹,

Vladova²

2021

TIM Review

View full text Add to dashboard Cite

learning, adapting, synthesizing, self-correction and the use of data for complex processing tasks". In addition, AI technology promises to provide deeper insights into learners' learning behaviours, reaction times, or emotions (Luckin et al., 2016;Holmes et al., 2019;Renz et al. 2020a). AI-driven tools can be categorized into two main areas: narrow AI/weak AI and general AI/strong AI. The former refers to an AI agent that is designed to solve one specific task, whereas the latter refers to an AI agent that is capable of solving multiple given problems irrespective of the task or domain. Almost all available educational tools comprise both narrow and general AI together, whereas a building a solely general/strong AI is unlikely to exist even in the future (Zawacki-Richter et al., 2019).Because the outcomes of these tools rely heavily on data produced in a specific task or domain, they affect people in several ways. For example, some are concerned about the use of private information, such as learner behaviours, abilities, and mental states while performing educational activities (Holmes et al., 2018). An increased

show abstract

Section: A Human-centered Ai Approach In Educationmentioning

confidence: 99%

Reinvigorating the Discourse on Human-Centered Artificial Intelligence in Educational Technologies

Renz¹,

Vladova²

2021

TIM Review

View full text Add to dashboard Cite

show abstract

“…As a result, in the present study, self-reported student performance (grades), self-reported student progress (improved knowledge and confidence), and self-reported use and usefulness were investigated. Intelligent tutoring systems (ITS) have been consistently shown to improve students' educational outcomes when used alone or combined with traditional instruction [44]. Likewise, intelligent tutoring system (ITS) technology, example follow-up tutors that can be generated without programming alone using cognitive tutor authoring tools (CTAT) [45].…”

Section: Research Articlementioning

confidence: 99%

Personal factors associated with satisfaction with tutoring in engineering students

et al. 2021

View full text Add to dashboard Cite

The tutoring of university students is very relevant for the achievement of engineering learning and personal factors are associated with satisfaction, establishing that the knowledge of these factors makes it possible to take measures that contribute to improve tutoring actions. The objective of the research was to establish the association of personal factors to satisfaction with tutoring in engineering students. A quantitative approach was chosen, the design was non-experimental and the association of personal factors with satisfaction with university tutoring. The population consisted of 800 engineering students and by means of probabilistic sampling with a 3% margin of error, the sample consisted of 458 students. The results showed that satisfaction with the university tutor is associated with the factors sex, age and college of origin, obtaining the Chi-square value x2 = 189.618a, 157.728a and 118.106a (p < 0.05); satisfaction with the university tutor associated with factors of sex, age and school of origin, obtained a Chi-square x2 = 189.618a, 157.728a and 118.106a (p < 0.05) and satisfaction with the organization and contents of the tutoring was associated with factors of sex, age and school of origin obtained a Chi-square x2 = 8.876a, 23.189a and 17.327a (p < 0.05). It is concluded that personal conditions are associated with satisfaction with tutoring, with the university tutor and with the organization and contents of tutoring in engineering students.

show abstract

“…As more examples are encountered the where-part of a skill tends to propose more and more candidate bindings, but it typically avoids overgeneralizing beyond what the examples so far entail. 6 The the AL where-learning mechanism described in [64] generalizes wherepart matching conditions by performing simple anti-unification. In this wherelearning mechanism, called 'AntiUnify', a set of conjunctive conditions are maintained that check for particular features or spatial relationships (i.e.…”

Section: Algorithmic-levelmentioning

confidence: 99%

“…In the multi-column addition example shown in Figure 5 these neighboring elements are eliminated upon generalization, however, in some circumstances these neighbor elements can remain and are essential to keeping where-learning from over-generalizing. For example, continuing with the example of multi-column addition, for skills that carry '1's across columns, neighboring elements help form an adjacency relationship between 6 Historically AL and SimStudent's where-learning mechanisms have been described as implementing version spaces [39] [36] [32] [64]. However, least general generalization is a better characterization since version spaces implement both general-to-specific and specific-to-general learning, but where-learning has historically only implemented the latter (section 6 describes why this is the case).…”

Section: Algorithmic-levelmentioning

confidence: 99%

Decomposed Inductive Procedure Learning

Weitekamp¹,

MacLellan²,

Harpstead³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Recent advances in machine learning have made it possible to train artificially intelligent agents that perform with super-human accuracy on a great diversity of complex tasks. However, the process of training these capabilities often necessitates millions of annotated examples-far more than humans typically need in order to achieve a passing level of mastery on similar tasks. Thus, while contemporary methods in machine learning can produce agents that exhibit superhuman performance, their rate of learning per opportunity in many domains is decidedly lower than human-learning. In this work we formalize a theory of Decomposed Inductive Procedure Learning (DIPL) that outlines how different forms of inductive symbolic learning can be used in combination to build agents that learn educationally relevant tasks such as mathematical, and scientific procedures, at a rate similar to human learners. We motivate the construction of this theory along Marr's concepts of the computational, algorithmic, and implementation levels of cognitive modeling, and outline at the computational-level six learning capacities that must be achieved to accurately model human learning. We demonstrate that agents built along the DIPL theory are amenable to satisfying these capacities, and demonstrate, both empirically and theoretically, that DIPL enables the creation of agents that exhibit human-like learning performance.

show abstract

An Interaction Design for Machine Teaching to Develop AI Tutors

Cited by 42 publications

References 15 publications

Reinvigorating the Discourse on Human-Centered Artificial Intelligence in Educational Technologies

Reinvigorating the Discourse on Human-Centered Artificial Intelligence in Educational Technologies

Personal factors associated with satisfaction with tutoring in engineering students

Decomposed Inductive Procedure Learning

Contact Info

Product

Resources

About