Explainable AI as a Social Microscope: A Case Study on Academic Performance

Sargsyan, Anahit; Karapetyan, Areg; Woon, Wei Lee; Alshamsi, Aamena

doi:10.1007/978-3-030-64583-0_24

Cited by 6 publications

(5 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Nemrégiben Vultureanu-Albişi és Bădică (2021) faalapú együttes (ensemble) osztályozókat használt a hallgatók kurzusokon nyújtott teljesítményének előrejelzésére nyilvánosan elérhető adathalmazok alapján, továbbá Nag recha és munkatársaihoz hasonlóan ők is alkal mazták a LIME algoritmust a modell lokális előrejelzéseinek értelmezésére (Nagrecha-Dillon-Chawla 2017). Sargsyan és munkatársai (Sargsyan et al 2020) klaszterelemzést végeztek a hasonló jellemzőkkel rendelkező diákok csoportjainak megtalálására. Először a klaszterezést a diákok előre jelzett tanulmányi átlaga alapján végezték el, majd a LIME által adott lokális előrejelzések magyarázatából származó súlyok alapján.…”

Section: Bevezetésunclassified

Lemorzsolódás előrejelzése személyre szabott értelmezhető gépi tanulási módszerek segítségével

Nagy

2023

ScientSec

View full text Add to dashboard Cite

Összefoglalás. A hallgatói lemorzsolódás az egyik legégetőbb probléma a felsőoktatásban. Ebben a munkában a lemorzsolódás előrejelzésén keresztül bemutatjuk, hogyan tudják segíteni a felsőoktatás résztvevőit a magyarázható mesterséges intelligencia (XAI) eszközök, mint például a permutációs fontosság, a parciális függőségi ábra és a SHAP. Végül pedig kitérünk a kutatás gyakorlati hasznosulásának lehetőségeire, például, hogy az egyéni előrejelzések magyarázata hogyan teszi lehetővé a személyre szabott beavatkozást. Az elemzések során azt találtuk, hogy a középiskolai tanulmányi átlag bír a legnagyobb prediktív erővel a végzés tényére vonatkozóan. Továbbá annak ellenére, hogy egy műszaki egyetem adatait elemeztük, azt találtuk, hogy a humán tárgyaknak is nagy inkrementális prediktív erejük van a végzés tényére vonatkozóan a reál tárgyakhoz képest. Summary. Delayed completion and student drop-out are some of the most critical problems in higher education, especially regarding STEM programs. A high drop-out rate induces both individual and economic loss, hence a detailed investigation of the main reasons for dropping out is warranted. Recently, there has been a lot of interest in the use of machine learning methods for the early detection of students at risk of dropping out. However, there has not been much debate on the use of interpretable machine learning (IML) and explainable artificial intelligence (XAI) technologies for dropout prediction. In this paper, we show how IML and XAI techniques can assist educational stakeholders in dropout prediction using data from the Budapest University of Technology and Economics. We demonstrate that complex black-box machine learning algorithms, for example CatBoost, are able to effectively detect at-risk student using only pre-enrollment achievement measures, but they lack interpretability. We demonstrate how the predictions can be explained both globally and locally using IML methods including permutation importance (PI), partial dependence plot (PDP), LIME, and SHAP values. Using global interpretations, we have found that the factor that has the greatest impact on academic performance is the high school grade point average, which measures general knowledge by taking into account grades in history, mathematics, Hungarian language and literature, a foreign language and a science subject. However, we also found that both mathematics and the subject of choice are among the most important variables, which suggests that program-specific knowledge is not negligible and complements general knowledge. We discovered that students are more likely to drop out if they do not start their university studies immediately after leaving secondary school. Using a partial dependence plot, we showed that humanities also have incremental predictive power, despite the fact that this analysis is based on data from a technical university. Finally, we also discuss the potential practical applications of our work, such as how the explanation of individual predictions allows for personalized interventions, for example by offering appropriate remedial courses and tutoring sessions. Our approach is unique in that we not only estimate the probability of dropping out, but also interpret the model and provide explanations for each prediction. As a result, this framework can be used in several fields. By predicting which majors they could be most successful in based on high school performance indicators, it might, for instance, assist high school students in selecting the appropriate programs at universities and hence this way it could be used for career assistance. Through the explanations of local predictions, the framework provided can also assist students in identifying the skills they need to develop to succeed in their university studies.

show abstract

Section: Bevezetésunclassified

Lemorzsolódás előrejelzése személyre szabott értelmezhető gépi tanulási módszerek segítségével

Nagy

2023

ScientSec

View full text Add to dashboard Cite

show abstract

“…The need for personalized explainable models is shown by [141] so that feedback from the intelligent tutoring system can increase the trust of students in feedback. LIME algorithm is the commonly used algorithm used to cluster the students based on their academic attainments [142] and to find the actual success indicators for specific students, localized models are important.…”

Section: Idmentioning

confidence: 99%

Closing the gap: exploring the untapped potential of machine learning in deaf students and hearing students’ academic performance

2023

IJATEE

View full text Add to dashboard Cite

Modern universities use several methods to improve the learning environment. This includes analyzing the academic performance of students and planning various intervention programs to support the students to overcome the difficulties they face. The performance prediction at the time of admission and in the subsequent years helps universities effectively plan the intervention strategies which in turn can help both management and educators benefit from the students' performance prediction plans [1]. Educational data mining (EDM) and machine learning (ML) methods are gaining popularity in the educational domain.*Author for correspondence Many applications using these techniques are used to analyse student performance in academics.The EDM and ML-based applications help instructors to identify struggling students earlier and take action to improve success and retention. Researchers are using these techniques to accelerate their research to uncover discoveries and insights. Though the role of ML and EDM is explored considerably in the education of hearing students, its benefits are not used properly in deaf education. Research suggests that deaf young adults are less likely than their hearing peers to live independently[2], are underemployed and underpaid [3], and are less likely to obtain an undergraduatelevel degree [4]. The opportunities for higher

show abstract

“…XAI has been used across a wide range of domains, including healthcare [175]- [181] (Section VI-A), finance [182]- [186] (Section VI-B), criminal justice [187]- [190] (Section VI-C) and other domains [191], [192] (Section VI-D). Given our focus on tabular data, we will discuss the applications that rely on such data.…”

Section: Applications Of Xaimentioning

confidence: 99%

“…In this context, the SHAP dependency plots were used to highlight any important non-linear relationships between the independent features and the dependent variables, thereby providing insights into the improvement of existing policies. Finally, we mention the work of Sargsyan et al [192], who classified students based on their LIME coefficient. This allowed them to identify the students who have similar academic attainment indicators, thereby providing a more nuanced view of the success indicators.…”

Section: Other Domainsmentioning

confidence: 99%

Explainable Artificial Intelligence for Tabular Data: A Survey

2021

View full text Add to dashboard Cite

Machine learning techniques are increasingly gaining attention due to their widespread use in various disciplines across academia and industry. Despite their tremendous success, many such techniques suffer from the "black-box" problem, which refers to situations where the data analyst is unable to explain why such techniques arrive at certain decisions. This problem has fuelled interest in Explainable Artificial Intelligence (XAI), which refers to techniques that can easily be interpreted by humans. Unfortunately, many of these techniques are not suitable for tabular data, which is surprising given the importance and widespread use of tabular data in critical applications such as finance, healthcare, and criminal justice. Also surprising is the fact that, despite the vast literature on XAI, there are still no survey articles to date that focus on tabular data. Consequently, despite the existing survey articles that cover a wide range of XAI techniques, it remains challenging for researchers working on tabular data to go through all of these surveys and extract the techniques that are suitable for their analysis. Our article fills this gap by providing a comprehensive and up-to-date survey of the XAI techniques that are relevant to tabular data. Furthermore, we categorize the references covered in our survey, indicating the type of the model being explained, the approach being used to provide the explanation, and the XAI problem being addressed. Our article is the first to provide researchers with a map that helps them navigate the XAI literature in the context of tabular data. INDEX TERMSBlack-box models, Explainable Artificial Intelligence, Machine Learning, Model interpretability VOLUME 4, 2016This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3116481, IEEE Access M. Sahakyan et al.: Explainable Artificial Intelligence for Tabular Data: A Survey• Decomposability, reflecting the degree to which each input, parameter, and calculation can be explained intuitively;• Algorithmic transparency, reflecting the degree to which the inner workings of the learning algorithm can be understood.For example, rule-based models [16] are considered transparent since they use a series of if-then rules that can easily be understood without the need for any further explanation. Unlike transparent models, black-box models are those that do not explain their predictions in a way that humans can understand [17]. Examples of black-box models include artificial neural networks [18] and gradient boosting [19]. Although black-box models are hard to interpret by humans, they tend to have higher prediction accuracy compared to their transparent counterparts. This trade-off between accuracy and transparency gives raise to the black-box explanation problem, which involves explaining the rationale behind the decisions made by black-box models. By providing such explanations, one can continue ...

show abstract

Explainable AI as a Social Microscope: A Case Study on Academic Performance

Cited by 6 publications

References 18 publications

Lemorzsolódás előrejelzése személyre szabott értelmezhető gépi tanulási módszerek segítségével

Lemorzsolódás előrejelzése személyre szabott értelmezhető gépi tanulási módszerek segítségével

Closing the gap: exploring the untapped potential of machine learning in deaf students and hearing students’ academic performance

Explainable Artificial Intelligence for Tabular Data: A Survey

Contact Info

Product

Resources

About