Proactive and reactive engagement of artificial intelligence methods for education: a review

Probably not a substantial claim: may not pass S3 (load-bearing).

The number of review articles surveying technical progress in AIEd has increased in the last decade.

“In this backdrop of emerging developments, the number of review articles surveying the technical progress in the AIEd discipline has also increased in the last decade (see Figure 4).”

• AssumptionRests on an assumption: The Google Scholar search and keyword strategy captured the relevant AIEd literature.

Probably not a substantial claim: may not pass S3 (load-bearing).

Closely linked with technical reviews with categorization are review articles that dive deep into developments within a particular sub-category associated with AIEd, such as AIEd in early childhood education and online higher education.

“Closely linked with (i) are review articles that dive deep into the developments within a particular sub-category associated with AIEd, such as AIEd in the context of early childhood education (Su and Yang, 2022) and online higher education (Ouyang F. et al., 2022).”

• cites Su and Yang, 2022 sourceNot checked yet

  Not checked in the free run.

• cites Ouyang F. et al., 2022Not checked yet

  Not checked in the free run.

• AssumptionRests on an assumption: The Google Scholar search and keyword strategy captured the relevant AIEd literature.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 4 address timetabling under proactive content design.

“| Proactive engagement | Content design | Timetabling | 4 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing).

Among the 195 reviewed articles, 9 address teacher-focused performance assessment under reactive performance assessment.

“| Reactive engagement | Performance assessment | Teacher-focused | 9 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 13 address interactive tutoring aids under reactive tutoring aids.

“| Reactive engagement | Tutoring aids | Interactive tutoring aids | 13 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 8 address personalized tutoring aids under reactive tutoring aids.

“| Reactive engagement | Tutoring aids | Personalized tutoring aids | 8 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 22 address content generation under proactive engagement.

“| Proactive engagement | Content generation | N/A | 22 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 15 address learning content design under proactive content design.

“| Proactive engagement | Content design | Learning content design | 15 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 5 address affect aware tutoring aids under reactive tutoring aids.

“| Reactive engagement | Tutoring aids | Affect aware tutoring aids | 5 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 76 address student-focused performance assessment under reactive performance assessment.

“| Reactive engagement | Performance assessment | Student-focused | 76 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing).

Among the 195 reviewed articles, 13 address performance prediction under reactive outcome prediction.

“| Reactive engagement | Outcome prediction | Performance prediction | 13 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 7 address learning style based tutoring aids under reactive tutoring aids.

“| Reactive engagement | Tutoring aids | Learning style based tutoring aids | 7 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 9 address student admission logistics under proactive engagement.

“| Proactive engagement | Student admission logistics | N/A | 9 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Among the 195 reviewed articles, 14 address drop-out prediction under reactive outcome prediction.

“| Reactive engagement | Outcome prediction | Drop-out prediction | 14 |”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

May combine two claims. The quote only partly states this claim.

The review covers 195 original research articles published between 2003 and 2022.

“We identified our selected corpus of 195 research articles through a multi-step process.”

• AssumptionRests on an assumption: The 195 selected articles are representative of original research on AI in education published between 2003 and 2022.

May combine two claims.

Earlier AIEd approaches used mostly static data (e.g., student records, administrative records, demographic information, surveys, and questionnaires), while later approaches used more real-time and high-volume data (e.g., student-teacher/peer-peer interaction data, click-stream information, web-browsing data).

“Whereas, earlier approaches used mostly static data (e.g., student records, administrative records, demographic information, surveys, and questionnaires), the use of more sophisticated algorithms necessitated (and in turn benefited from) more real-time and high-volume data (e.g., student-teacher/peer-peer interaction data, click-stream information, web-browsing data).”

May combine two claims.

In all categories, to different extents, there is a distinct shift from rule-based and statistical approaches to classical ML to deep learning methods, and from handcrafted features to automatically extracted features.

“In all categories, to different extents, we see a distinct shift from rule-based and statistical approaches to classical ML to deep learning methods, and from handcrafted features to automatically extracted features.”

May combine two claims.

The growing interest in AIEd can be inferred from Figures 2 and 4, which show how both the count of technical contributions and the count of review articles on the topic have increased over the past two decades.

“The growing interest in AIEd can be inferred from Figures 2, 4 which show how both the count of technical contributions and the count of review articles on the topic have increased over the past two decades.”

Probably not a substantial claim: may not pass S4 (based).

Even though data-hungry models like deep neural networks have grown in popularity across almost all categories discussed here, AIEd often suffers from the availability of sufficient labeled data to train such systems.

“Even though data-hungry models like deep neural networks have grown in popularity across almost all categories discussed here, AIEd often suffers from the availability of sufficient labeled data to train such systems.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Through this review, we identified the paradigm shift over the past 20 years in formulating computational models (i.e., choice of algorithms, choice of features etc.) and training them (i.e., choice of data)-we are indeed increasingly leaning toward sophisticated yet explainable frameworks.

“Through this review, we identified the paradigm shift over the past 20 years in formulating computational models (i.e., choice of algorithms, choice of features etc.) and training them (i.e., choice of data)—we are indeed increasingly leaning toward sophisticated yet explainable frameworks.”

May combine two claims.

Based on our survey, while in some parts of the world we have seen great momentum in making AIEd a part and parcel of the education sector, in other parts of the world this progress is stymied by inadequate access to necessary infrastructure and human resources.

“Based on our survey, we have observed that while in some parts of the world we have seen great momentum in making AIEd a part and parcel of the education sector, in other parts of the world this progress is stymied by inadequate access to necessary infrastructure and human resources.”

May combine two claims.

AIEd research as it stands today only augments the role of the teacher, and does not eliminate it.

“AIEd research as it stands today only augments the role of the teacher, and does not eliminate it.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

In the proactive engagement phase, AI-based algorithms can be leveraged to determine student admission logistics, design curricula and schedules, and create course content.

“For instance, in the proactive engagement phase, AI-based algorithms can be leveraged to determine student admission logistics, design curricula and schedules, and create course content.”

May combine two claims.

Zhao et al. (2020) used features extracted from application materials of students as well as how they performed in the program of study to predict an incoming applicant's potential performance and identify students best suited for the program.

“Zhao et al. (2020) used features extracted from application materials of students as well as how they performed in the program of study to predict an incoming applicant's potential performance and identify students best suited for the program.”

• cites Zhao et al., 2020Not checked yet

  Not checked in the free run.

May combine two claims.

GRADE did not make the final admission decision but reduced the number of full application reviews and review time per application by experts.

“While GRADE did not make the final admission decision, it reduced the number of full application reviews as well as review time per application by experts.”

• cites Waters and Miikkulainen, 2014Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Integrating computer-based technology and digital learning tools can enhance the learning experience for students and knowledge delivery process for educators.

“Integrating computer-based technology and digital learning tools can enhance the learning experience for students and knowledge delivery process for educators (Lin et al., 2017; Mei et al., 2019).”

• cites Lin et al., 2017Not checked yet

  Not checked in the free run.

• cites Mei et al., 2019Not checked yet

  Not checked in the free run.

May combine two claims.

The proactive phase primarily uses historical data points or pre-existing estimates of available resources and expectations about learning outcomes, whereas the reactive phase uses a growing pool of data points from the ongoing learning process and can therefore be more adaptive and initiate faster pedagogical interventions.

“While the former primarily makes use of historical data points or pre-existing estimates of available resources and expectations about learning outcomes, the latter has at its disposal a growing pool of data points from the currently ongoing learning process, and can therefore be more adaptive and initiate faster pedagogical interventions to changing scopes and requirements.”

May combine two claims.

In the reactive engagement phase, AI-based methods can be used for designing intelligent tutoring systems, performance assessment, and prediction of student outcomes.

“On the other hand, in the reactive engagement phase, AI-based methods can be used for designing intelligent tutoring systems (ITS), performance assessment, and prediction of student outcomes (see Figure 1).”

Classified as probably not a substantial claim.

There are two distinct phases of engaging AI in the end-to-end education process: proactive engagement and reactive engagement.

“To this end, we can clearly identify that there are two distinct phases of engaging AI in the end-to-end education process.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Using AI-driven platforms and tools to support students, educators, and policy-makers in education appears to be more feasible than ever.

“With this parallel ongoing progress, using AI-driven platforms and tools to support students, educators, and policy-makers in education appears to be more feasible than ever.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

This review article is the first to present a distinction between proactive and reactive involvement of AI in education along with a granular review of significant research questions in each category.

“To the best of our knowledge, such distinction between proactive and reactive involvement of AI in education along with an granular review of significant research questions in each category is presented for the first time through this paper (see schematic in Figure 1).”

• AssumptionRests on an assumption: The Google Scholar search and keyword strategy captured the relevant AIEd literature.

Probably not a substantial claim: may not pass S3 (load-bearing).

Integrating computer-based technology and digital learning tools can help accelerate administrative tasks related to education.

“It can also help accelerate administrative tasks related to education (Ahmad et al., 2020).”

• cites Ahmad et al., 2020Not checked yet

  Not checked in the free run.

May combine two claims.

Emelianov et al. (2020) showed that demographic parity mechanisms like group-specific admission thresholds increase the utility of the selection process in such systems in addition to improving its fairness.

“Emelianov et al. (2020) showed that demographic parity mechanisms like group-specific admission thresholds increase the utility of the selection process in such systems in addition to improving its fairness.”

• cites Emelianov et al., 2020Not checked yet

  Not checked in the free run.

May combine two claims.

QDP was historically estimated by pretesting on students or from expert ratings, which are expensive, time-consuming, subjective, and often vulnerable to leakage or exposure.

“QDP was historically estimated by pretesting on students or from expert ratings, which are expensive, time-consuming, subjective, and often vulnerable to leakage or exposure (Benedetto et al., 2022).”

• cites Benedetto et al., 2022Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Mathematical word problems (MWPs) have witnessed significant research interest in automatic generation.

“In this context, mathematical word problems (MWPs)—an established way of inculcating math modeling skills in K-12 education—have witnessed significant research interest.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Machine learning has often been used in conjunction with mathematical techniques to obtain better performing algorithms for timetabling.

“Having said that, it must be noted that machine learning has often been used in conjunction with such mathematical techniques to obtain better performing algorithms.”

• cites Kenekayoro, 2019Not checked yet

  Not checked in the free run.

• cites Obit et al., 2011Not checked yet

  Not checked in the free run.

• cites Goh et al., 2019Not checked yet

  Not checked in the free run.

May combine two claims.

Reinforcement learning has been used to select low-level heuristics in hyper-heuristic approaches or to obtain a suitable search neighborhood in mathematical optimization problems for timetabling.

“Reinforcement learning has been used to select low-level heuristics in hyper-heuristic approaches (Obit et al., 2011; Özcan et al., 2012) or to obtain a suitable search neighborhood in mathematical optimization problems (Goh et al., 2019).”

• cites Obit et al., 2011Not checked yet

  Not checked in the free run.

• cites Özcan et al., 2012Not checked yet

  Not checked in the free run.

• cites Goh et al., 2019Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

A majority of data-driven QDP approaches belong to language learning and medicine.

“AlKhuzaey et al. (2021) points out that a majority of data-driven QDP approaches belong to language learning and medicine, possibly spurred on by the existence of a large number of international and national-level standardized language proficiency tests and medical licensing exams.”

• cites AlKhuzaey et al., 2021Not checked yet

  Not checked in the free run.

May combine two claims.

Content selection for questions was typically done using statistical features (sentence length, word/sentence position, word frequency, noun/pronoun count, presence of superlatives, etc.) or NLP techniques such as syntactic or semantic parsing, named entity recognition, and topic modeling.

“Content selection for questions was typically done using different statistical features (sentence length, word/sentence position, word frequency, noun/pronoun count, presence of superlatives, etc.) (Agarwal and Mannem, 2011) or NLP techniques such as syntactic or semantic parsing (Heilman, 2011; Lindberg et al., 2013), named entity recognition (Kalady et al., 2010) and topic modeling (Majumder and”

• cites Agarwal and Mannem, 2011Not checked yet

  Not checked in the free run.

• cites Heilman, 2011Not checked yet

  Not checked in the free run.

• cites Kalady et al., 2010Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Following the success of neural language models and large-scale machine reading comprehension datasets, question generation was later framed as a sequence-to-sequence learning problem that directly maps a sentence (or the entire passage containing the sentence) to a question.

“Following the success of neural language models and concurrent with the release of large-scale machine reading comprehension datasets (Nguyen et al., 2016; Rajpurkar et al., 2016), question generation was later framed as a sequence-to-sequence learning problem that directly maps a sentence (or the entire passage containing the sentence) to a question (Du et al., 2017; Zhao et al., 2018; Kim et al.”

• cites Du et al., 2017Not checked yet

  Not checked in the free run.

• cites Zhao et al., 2018Not checked yet

  Not checked in the free run.

• cites Kim et al., 2019Not checked yet

  Not checked in the free run.

May combine two claims.

Attention strategies have been used to model the difficulty contribution of each sentence in reading problems or to model recall and confusion in QDP.

“Attention strategies have been used to model the difficulty contribution of each sentence in reading problems (Huang et al., 2017) or to model recall (how hard it is to recall the knowledge assessed by the question) and confusion (how hard it is to separate the correct answer from distractors) in Qiu et al. (2019).”

• cites Huang et al., 2017Not checked yet

  Not checked in the free run.

• cites Qiu et al., 2019Not checked yet

  Not checked in the free run.

May combine two claims.

Automatic question generation (AQG) from teaching material can be used to improve learning and comprehension of students, assess information retention from the material, and aid teachers in adding supplementary material from external sources without the time-intensive process of authoring assessments.

“Automatic question generation (AQG) from teaching material can be used to improve learning and comprehension of students, assess information retention from the material and aid teachers in adding Supplementary material from external sources without the time-intensive process of authoring assessments from them.”

May combine two claims.

Automatically generating problems that are similar to a given problem in terms of difficulty level can greatly benefit teachers in setting individualized practice problems to avoid plagiarism and ensure fair evaluation, and enables students to be exposed to as many diverse training exercises as needed to master underlying concepts.

“Automatically generating problems that are similar to a given problem in terms of difficulty level, can greatly benefit teachers in setting individualized practice problems to avoid plagiarism and still ensure fair evaluation (Ahmed et al., 2013). It also enables the students to be exposed to as many (and diverse) training exercises as needed in order to master the underlying concepts (Keller, 202”

• cites Ahmed et al., 2013Not checked yet

  Not checked in the free run.

• cites Keller, 2021Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Question construction traditionally adopted rule-based methods like transformation-based approaches or template-based approaches.

“The actual question construction, on the other hand, traditionally adopted rule-based methods like transformation-based approaches (Varga and Ha, 2010) or template-based approaches (Mostow and Chen, 2009).”

• cites Varga and Ha, 2010Not checked yet

  Not checked in the free run.

• cites Mostow and Chen, 2009Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

AQG systems have also been developed for structured databases, images, and videos, and are typically evaluated by experts on the quality of generated questions in terms of relevance, grammatical and semantic correctness, usefulness, and clarity.

“While text is the most common type of input in AQG, such systems have also been developed for structured databases (Jouault and Seta, 2013; Indurthi et al., 2017), images (Mostafazadeh et al., 2016), and videos (Huang et al., 2014), and are typically evaluated by experts on the quality of generated questions in terms of relevance, grammatical, and semantic correctness, usefulness, clarity etc.”

• cites Jouault and Seta, 2013Not checked yet

  Not checked in the free run.

• cites Mostafazadeh et al., 2016 sourceNot checked yet

  Not checked in the free run.

• cites Huang et al., 2014Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

AI has been widely adopted to generate and improve learning content prior to the start of the learning process.

“AI has been widely adopted to generate and improve learning content prior to the start of the learning process, as discussed in this section.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Timetabling design considerations (e.g., number of courses per semester, lectures per day, free time-slots per day) have a noticeable impact on student attendance behavior and academic performance.

“Timetabling not only ensures proper resource allocation, its design considerations (e.g., number of courses per semester, number of lectures per day, number of free time-slots per day) have noticeable impact on student attendance behavior and academic performance (Larabi-Marie-Sainte et al., 2021).”

• cites Larabi-Marie-Sainte et al., 2021Not checked yet

  Not checked in the free run.

May combine two claims.

A number of Automatic Essay Scoring (AES) and Automatic Short Answer Grading (ASAG) systems have been developed to reliably evaluate compositions produced by learners in response to a given prompt, and are typically trained on a large set of written samples pre-scored by expert raters.

“In the context of written prose, a number of Automatic Essay Scoring (AES) and Automatic Short Answer Grading (ASAG) systems have been developed to reliably evaluate compositions produced by learners in response to a given prompt, and are typically trained on a large set of written samples pre-scored by expert raters (Shermis and Burstein, 2003; Dikli, 2006).”

• cites Shermis and Burstein, 2003Not checked yet

  Not checked in the free run.

• cites Dikli, 2006Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

AQA, MRC, and automatic quantitative reasoning problem solvers have achieved impressive breakthroughs with large language models and are widely regarded as a stepping-stone toward Artificial General Intelligence (AGI).

“These have achieved impressive breakthroughs with the research into large language models and are widely regarded in the larger narrative as a stepping-stone toward Artificial General Intelligence (AGI), since they require sophisticated natural language understanding and logical inferencing capabilities.”

May combine two claims.

Approaches proposed to overcome the bottleneck of labeled code samples include learning question-independent features from code samples or zero-shot learning using human-in-the-loop rubric sampling.

“Approaches proposed to overcome this issue include learning question-independent features from code samples (Singh et al., 2016; Tarcsay et al., 2022) or zero-shot learning using human-in-the-loop rubric sampling (Wu et al., 2019).”

• cites Singh et al., 2016Not checked yet

  Not checked in the free run.

• cites Tarcsay et al., 2022Not checked yet

  Not checked in the free run.

• cites Wu et al., 2019Not checked yet

  Not checked in the free run.

May combine two claims.

For categorization per the Felder and Silverman learning style model, count of student visits to different sections of the e-learning platform are found to be more informative; for categorization per the Kolb learning model, student performance and student preference features were found to be more relevant.

“While for categorization per the Felder and Silverman learning style model, count of student visits to different sections of the e-learning platform are found to be more informative (Bernard et al., 2015; Bajaj and Sharma, 2018), for categorization per the Kolb learning model, student performance, and student preference features were found to be more relevant.”

• cites Bernard et al., 2015Not checked yet

  Not checked in the free run.

• cites Bajaj and Sharma, 2018Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Customizing course content according to learning styles of students can lead to better end outcomes.

“Yet another perspective in the domain of ITS pertains to customizing course content according to learning styles of students for better end outcomes.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Knowledge tracing models knowledge development in students in order to predict their ability to answer the next problem correctly given their current mastery level of knowledge concepts.

“An effective way of monitoring the learning progress of students is through knowledge tracing, which models knowledge development in students in order to predict their ability to answer the next problem correctly given their current mastery level of knowledge concepts.”

May combine two claims.

Doroudi (2019) presents a systematic review of RL-induced instructional policies that were evaluated on students, and concludes that over half outperform all baselines they were tested against.

“Doroudi (2019) presents a systematic review of RL-induced instructional policies that were evaluated on students, and concludes that over half outperform all baselines they were tested against.”

• cites Doroudi, 2019Not checked yet

  Not checked in the free run.

May combine two claims.

Automatic evaluation reduces the burden on teaching assistants and graders, removes grader-to-grader variability in assessment, and helps accelerate the learning process for students by providing real-time feedback.

“But automatic evaluation remains an important problem not only because it reduces the burden on teaching assistants and graders, but also removes grader-to-grader variability in assessment and helps accelerate the learning process for students by providing real-time feedback (Srikant and Aggarwal, 2014).”

• cites Srikant and Aggarwal, 2014Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The applicability and usefulness of AQA, MRC, and automatic quantitative reasoning problem solvers in educational settings remains to be seen.

“However, their applicability and usefulness in educational settings remains to be seen.”

Probably not a substantial claim: may not pass S3 (load-bearing).

AI-based evaluation tools are heavily used in computer science education, particularly programming, due to its inherent structure and logic.

“AI-based evaluation tools are also heavily used in computer science education, particularly programming, due to its inherent structure and logic.”

May combine two claims.

AES systems are increasingly being adopted into Automated Writing Evaluation (AWE) systems that provide formative feedback along with (or instead of) final scores and therefore have greater pedagogical usefulness.

“Moreover, feedback being a critical aspect of essay drafting and revising, AES systems are increasingly being adopted into Automated Writing Evaluation (AWE) systems that provide formative feedback along with (or instead of) final scores and therefore have greater pedagogical usefulness (Hockly, 2019).”

• cites Hockly, 2019Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Scoring submitted work and providing feedback on similar scales remains a difficult problem.

“While technological developments have made it easier to provide content to learners at scale, scoring their submitted work and providing feedback on similar scales remains a difficult problem.”

May combine two claims.

Knowledge tracing benefits students by identifying areas they need to work on, and educators in designing targeted exercises, personalized learning recommendations and adaptive teaching strategies.

“This not only benefits the students by identifying areas they need to work on, but also the educators in designing targeted exercises, personalized learning recommendations and adaptive teaching strategies (Liu et al., 2019).”

• cites Liu et al., 2019Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Handling of non-textual content as in images, mathematical equations or code snippets to learn richer embedding representations of questions or knowledge concepts remains relatively unexplored in the domain of knowledge tracing.

“However as pointed out in Abdelrahman et al. (2022), handling of non-textual content as in images, mathematical equations or code snippets to learn richer embedding representations of questions or knowledge concepts remains relatively unexplored in the domain of knowledge tracing.”

• cites Abdelrahman et al., 2022Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Previous research in education (Jackson and McNamara, 2013) has pointed out that when a student is actively interacting with the educator or the course contents, the student stays engaged in the learning process for a longer duration.

“Previous research in education (Jackson and McNamara, 2013) has pointed out that when a student is actively interacting with the educator or the course contents, the student stays engaged in the learning process for a longer duration.”

• cites Jackson and McNamara, 2013Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S4 (based).

The sequence in which a student reviews course topics plays an important role in their mastery of a concept.

“The sequence in which a student reviews course topics plays an important role in their mastery of a concept.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Since a majority of evaluation material are in textual format, NLP-based models in particular have a major presence in AI-driven performance assessment and monitoring.

“Since a majority of evaluation material are in textual format, NLP-based models in particular have a major presence in this domain.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Deep knowledge tracing (DKT) was one of the first models which used recurrent neural network architectures for modeling the latent knowledge state along with its temporal dynamics to predict future performance.

“Deep knowledge tracing (DKT) was one of the first such models which used recurrent neural network architectures for modeling the latent knowledge state along with its temporal dynamics to predict future performance (Piech et al., 2015a).”

• cites Piech et al., 2015aNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

A more useful direction for automatic grading measures similarities between abstract representations (control flow graphs, system dependence graphs) of the student's program and correct implementations of the program.

“A more useful direction measures similarities between abstract representations (control flow graphs, system dependence graphs) of the student's program and correct implementations of the program (Wang et al., 2007; Vujošević-Janičić et al., 2013) for automatic grading.”

• cites Wang et al., 2007Not checked yet

  Not checked in the free run.

• cites Vujošević-Janičić et al., 2013Not checked yet

  Not checked in the free run.

May combine two claims.

Traditional approaches for automated grading of source codes such as test-case based assessments and assessments using code metrics, while simple, are neither robust nor effective at evaluating program quality.

“Traditional approaches for automated grading of source codes such as test-case based assessments (Douce et al., 2005) and assessments using code metrics (e.g., lines of code, number of variables, number of statements), while simple, are neither robust nor effective at evaluating program quality.”

• cites Douce et al., 2005Not checked yet

  Not checked in the free run.

May combine two claims.

AES systems have been developed for holistic scoring and for fine-grained evaluation along specific dimensions of essay quality, such as organization, prompt-adherence, thesis clarity, argument strength, and thesis strength.

“Such systems have been developed not only to provide holistic scoring (assessing essay quality with a single score), but also for more fine-grained evaluation by providing scoring along specific dimensions of essay quality, such as organization (Persing et al., 2010), prompt-adherence (Persing and Ng, 2014), thesis clarity (Persing and Ng, 2013), argument strength (Persing and Ng, 2015), and”

• cites Persing et al., 2010Not checked yet

  Not checked in the free run.

• cites Persing and Ng, 2014Not checked yet

  Not checked in the free run.

• cites Ke et al., 2019Not checked yet

  Not checked in the free run.

May combine two claims.

The review identified four hypotheses currently receiving much attention from the research community in ITS designs: emphasis on tutor-tutee interaction, emphasis on personalization, inclusion of affect and emotion, and consideration of specific learning styles.

“We primarily identified four hypotheses that are currently receiving much attention from the research community—emphasis on tutor-tutee interaction, emphasis of personalization, inclusion of affect and emotion, and consideration of specific learning styles.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Affect and behavioral state of a student can generally be inferred from sensor data that tracks minute physical movements of the student (eyegaze, facial expression, posture etc.).

“Affect and behavioral state of a student can generally be inferred from sensor data that tracks minute physical movements of the student (eyegaze, facial expression, posture etc.).”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

AI-driven performance assessment and monitoring tools have been widely developed for both learners and educators.

“AI-driven performance assessment and monitoring tools have been widely developed for both learners and educators.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Xu et al. (2017) proposed that a student's performance or when the student anticipates graduation should be predicted progressively using an ensemble machine learning method over the duration of the student's tenure as the academic state of the student is ever-evolving and can be traced through their student records.

“Xu et al. (2017) proposed that a student's performance or when the student anticipates graduation should be predicted progressively (using an ensemble machine learning method) over the duration of the student's tenure as the academic state of the student is ever-evolving and can be traced through their student records.”

• cites Xu et al. (2017)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Researchers have aimed to improve the performance of performance predictors by including relevant information such as student engagement, interactions, role of external incentives, and previous performance records.

“Researchers have aimed to improve the performance of the predictors by including relevant information such as student engagement, interactions (Ramesh et al., 2013; Bydžovská, 2016), role of external incentives (Jiang et al., 2014), and previous performance records (Tamhane et al., 2014).”

• cites Ramesh et al., 2013Not checked yet

  Not checked in the free run.

• cites Bydžovská, 2016Not checked yet

  Not checked in the free run.

• cites Jiang et al., 2014Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Foltỳnek et al. (2019) classifies plagiarism forms according to an increasing order of obfuscation level, from verbatim and near-verbatim copying to translation, paraphrasing, idea-preserving plagiarism, and ghostwriting.

“Foltỳnek et al. (2019) in their review paper on academic plagiarism detection in text (e.g., essays, reports, research papers) classifies plagiarism forms according to an increasing order of obfuscation level, from verbatim and near-verbatim copying to translation, paraphrasing, idea-preserving plagiarism, and ghostwriting.”

• cites Foltỳnek et al. (2019)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

In computer science education, source code plagiarism can be classified based on increasing levels of obfuscation.

“In computer science education where programming assignments are given to evaluate students, source code plagiarism can also been classified based on increasing levels of obfuscation (Faidhi and Robinson, 1987).”

• cites Faidhi and Robinson, 1987Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Waheed et al. (2020) used deep feed-forward neural networks and split the problem of predicting student grade into multiple binary classification problems viz., Pass-Fail, Distinction-Pass, Distinction-Fail, Withdrawn-Pass.

“For example, Waheed et al. (2020) used deep feed-forward neural networks and split the problem of predicting student grade into multiple binary classification problems viz., Pass-Fail, Distinction-Pass, Distinction-Fail, Withdrawn-Pass.”

• cites Waheed et al. (2020)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S4 (based).

Such a priori understanding of a student's outcome allows for designing effective intervention strategies.

“Such apriori understanding of a student's outcome allows for designing effective intervention strategies.”

May combine two claims.

Several works use a combination of student ratings and attributes related to the course and the instructor to predict instructor performance and investigate factors affecting learning outcomes.

“Several works use a combination of student ratings and attributes related to the course and the instructor to predict instructor performance and investigate factors affecting learning outcomes (Mardikyan and Badur, 2011; Ahmed et al., 2016; Abunasser et al., 2022) .”

• cites Mardikyan and Badur, 2011Not checked yet

  Not checked in the free run.

• cites Ahmed et al., 2016Not checked yet

  Not checked in the free run.

• cites Abunasser et al., 2022Not checked yet

  Not checked in the free run.

May combine two claims.

Plagiarism detection methods have been developed for increasingly complex types of plagiarism, and widely adopt NLP and ML-based techniques for each.

“In a similar fashion, plagiarism detection methods have been developed for increasingly complex types of plagiarism, and widely adopt NLP and ML-based techniques for each (Foltỳnek et al., 2019).”

• cites Foltỳnek et al., 2019Not checked yet

  Not checked in the free run.

May combine two claims.

Educators can benefit from automatic prediction of students' performance and automatic identification of students at risk of course non-completion.

“On the other hand, educators can also benefit from automatic prediction of students' performance and automatic identification of students at risk of course non-completion.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Tsiakmaki et al. (2020) analyzed if transfer learning can be used to accurately predict student performance.

“Tsiakmaki et al. (2020) analyzed if transfer learning (i.e., pre-training neural networks on student data on a different course) can be used to accurately predict student performance.”

• cites Tsiakmaki et al. (2020)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

AI itself makes plagiarism detection an uphill battle.

“It is worth noting that AI itself makes plagiarism detection an uphill battle.”

May combine two claims.

Automatic prediction of students' performance and identification of at-risk students can be accomplished by monitoring students' patterns of engagement with the course material in association with their demographic information.

“This can be accomplished by monitoring students' patterns of engagement with the course material in association with their demographic information.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Syntax-based plagiarism detection methods rely on Part-of-speech tagging, frequency of PoS tags, or comparison of syntactic trees.

“Syntax-based methods rely on Part-of-speech (PoS) tagging (Gupta et al., 2014), frequency of PoS tags (Hürlimann et al., 2015), or comparison of syntactic trees (Tschuggnall and Specht, 2013).”

• cites Gupta et al., 2014Not checked yet

  Not checked in the free run.

• cites Hürlimann et al., 2015Not checked yet

  Not checked in the free run.

• cites Tschuggnall and Specht, 2013Not checked yet

  Not checked in the free run.

May combine two claims.

With the increasing prevalence of easily accessible large language models like InstructGPT and ChatGPT that are capable of producing natural-sounding essays and short answers, and even working code snippets in response to a text prompt, it is now easier than ever for dishonest learners to misuse such systems for authoring assignments, projects, research papers or online exams.

“With the increasing prevalence of easily accessible large language models like InstructGPT (Ouyang L. et al., 2022) and ChatGPT (Blog, 2022) that are capable of producing natural-sounding essays and short answers, and even working code snippets in response to a text prompt, it is now easier than ever for dishonest learners to misuse such systems for authoring assignments, projects, research papers”

• cites Ouyang L. et al., 2022Not checked yet

  Not checked in the free run.

• cites Blog, 2022Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S4 (based).

Poor performance in any assessment cannot be generalized into a course non-completion.

“Note that we discuss these separately as poor performance in any assessment cannot be generalized into a course non-completion.”

Probably not a substantial claim: may not pass S3 (load-bearing).

He et al. (2015) addresses the concern that weekly prediction of probability of a student dropping out might have wide variance by including smoothing techniques.

“He et al. (2015) addresses the concern that weekly prediction of probability of a student dropping out might have wide variance by including smoothing techniques.”

• cites He et al. (2015)Not checked yet

  Not checked in the free run.

May combine two claims.

MOOCs have a significantly high drop-out rate, with a disparity between the number of students who sign up for a course and the number who actually complete the course.

“While MOOCs promise the democratization of education, one of the biggest concerns with MOOCs is the disparity between the number of students who sign up for a course versus the number of students who actually complete the course—the drop-out rate in MOOCs is significantly high (Hollands and Kazi, 2018; Reich and Ruipérez-Valiente, 2019).”

• cites Hollands and Kazi, 2018Not checked yet

  Not checked in the free run.

• cites Reich and Ruipérez-Valiente, 2019Not checked yet

  Not checked in the free run.

May combine two claims.

The cost of deploying and operating AIEd tools often alienates communities with limited means, thereby widening the gap in access to education.

“The paradox here is that the cost of deploying and operating AIEd tools often alienates communities with limited means thereby widening the gap in access to education.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Salas-Pilco et al. (2022) also outlines the technological, pedagogical, and socio-cultural barriers for AIEd in inclusive education.

“Salas-Pilco et al. (2022) also outlines the technological, pedagogical, and socio-cultural barriers for AIEd in inclusive education.”

• cites Salas-Pilco et al. (2022)Not checked yet

  Not checked in the free run.

May combine two claims.

Because AIEd platforms are privately owned, there is little transparency and regulations regarding their development and operations.

“However, as these platforms are privately owned, there is little transparency and regulations regarding their development and operations.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Lakkaraju et al. (2015) recommends assigning a risk-score per student rather than a binary label because resources to intervene might be limited.

“On the other hand, as resources to intervene might be limited, Lakkaraju et al. (2015) recommends assigning a risk-score per student rather than a binary label.”

• cites Lakkaraju et al. (2015)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The COVID-19 pandemic has emphasized the necessity of MOOCs.

“The COVID-19 pandemic has only emphasized this necessity (Purkayastha and Sinha, 2021).”

• cites Purkayastha and Sinha, 2021Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The number of technical contributions in 2021 and 2022 might have fallen in part due to pandemic-related indefinite school closures and shift to alternate learning models.

“It is to be noted that the number of technical contributions in 2021 and 2022 (assuming our sample of reviewed articles is representative of the population) might have fallen in part due to pandemic-related indefinite school closures and shift to alternate learning models.”

• AssumptionThe authors state this as an assumption: The sample of reviewed articles is representative of the population of AIEd technical contributions.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

The causes for dropping out of a course or a degree program can be diverse, but early prediction of it allows administrative staff and educators to intervene.

“The causes for dropping out of a course or a degree program can be diverse, but early prediction of it allows administrative staff and educators to intervene.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Chui et al. (2020) used a generative adversarial network based architecture to address the challenges of low volume of training data in alternative learning paradigms such as supportive learning.

“Chui et al. (2020) used a generative adversarial network based architecture, to address the challenges of low volume of training data in alternative learning paradigms such as supportive learning.”

• cites Chui et al. (2020)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Several recent works have investigated the role of pandemic-driven remote and hybrid instruction in widening gaps in educational achievements by race, poverty level, and gender.

“Several recent works have investigated the role of pandemic-driven remote and hybrid instruction in widening gaps in educational achievements by race, poverty level, and gender (Halloran et al., 2021; UNESCO, 2021; Goldhaber et al., 2022).”

• cites Halloran et al., 2021Not checked yet

  Not checked in the free run.

• cites UNESCO, 2021Not checked yet

  Not checked in the free run.

• cites Goldhaber et al., 2022Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Brooks et al. (2015) considers the level of activity of a student in bins of time during a semester as binary features (active vs. inactive) and then uses these sequences as n-grams to predict drop-out.

“Brooks et al. (2015) considers the level of activity of a student in bins of time during a semester as a binary features (active vs. inactive) and then uses these sequences as n-grams to predict drop-out.”

• cites Brooks et al. (2015)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Remote learning requires a level of tech-savviness on the part of students and teachers alike, which might be less prevalent in people with learning disabilities.

“Moreover, remote learning requires a level of tech-savviness on the part of students and teachers alike, which might be less prevalent in people with learning disabilities.”

May combine two claims.

Most AIEd researchers are concentrated in the technologically advanced parts of the world, and most AIEd platforms and applications are owned currently by the private sector.

“As pointed out by Pedro et al. (2019), besides most AIEd researchers being concentrated in the technologically advanced parts of the world, most AIEd platforms and applications are owned currently by the private sector.”

• cites Pedro et al. (2019)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Garg and Sharma (2020) outlines the different ways AI is used in special need education for development of adaptive and inclusive pedagogies.

“In this context, Garg and Sharma (2020) outlines the different ways AI is used in special need education for development of adaptive and inclusive pedagogies.”

• cites Garg and Sharma (2020) sourceNot checked yet

  Not checked in the free run.

May combine two claims.

A widespread transition to remote learning necessitates access to proper infrastructure (electricity, internet connectivity, and smart electronic devices) as well as resources (learning material, textbooks, educational softwares, etc.), which create barriers for low-income groups.

“A widespread transition to remote learning necessitates access to proper infrastructure (electricity, internet connectivity, and smart electronic devices that can support video conferencing apps and basic file sharing) as well as resources (learning material, textbooks, educational softwares, etc.), which create barriers for low-income groups (Muñoz-Najar et al., 2021).”

• cites Muñoz-Najar et al., 2021Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Dien et al. (2020) proposed extensive data pre-processing using min-max scaler, quantile transformation, etc. before passing the data in a deep-learning model such as one-dimensional convolutional network (CN1D) or recurrent neural networks.

“Dien et al. (2020) proposed extensive data pre-processing using min-max scaler, quantile transformation, etc. before passing the data in a deep-learning model such as one-dimensional convolutional network (CN1D) or recurrent neural networks.”

• cites Dien et al. (2020)Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Salas-Pilco et al. (2022) reviews the different ways in which AI positively impacts education of minority students, e.g., through facilitating performance/engagement improvement, student retention, student interest in STEM/STEAM fields, etc.

“Salas-Pilco et al. (2022) reviews the different ways in which AI positively impacts education of minority students, e.g., through facilitating performance/engagement improvement, student retention, student interest in STEM/STEAM fields, etc.”

• cites Salas-Pilco et al. (2022)Not checked yet

  Not checked in the free run.

May combine two claims.

The pandemic era saw an increasing adoption of video conferencing softwares and social media platforms for knowledge delivery, combined with more asynchronous formats of learning.

“As lockdowns and social distancing mandated a hastened transition to fully virtual delivery of educational content, the pandemic era saw an increasing adoption of video conferencing softwares and social media platforms for knowledge delivery, combined with more asynchronous formats of learning.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The ethical and legal implications for large-scale adoption of AI for education is a topic of active debate.

“The ethical and legal implications for large-scale adoption of AI for education is also a topic of active debate (Holmes and Porayska-Pomsta, 2022).”

• cites Holmes and Porayska-Pomsta, 2022Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

For the foreseeable future, we would need a human in the loop to provide feedback and ensure proper daily usage of these tools.

“Furthermore, for the foreseeable future, we would need a human in the loop to provide feedback and ensure proper daily usage of these tools.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The main concerns regarding AI in education are related to data privacy and security, job security, ethics etc. post adoption of AI in this sector.

“The main concerns are related to data privacy and security, job security, ethics etc. post adoption of AI in this sector.”