Background: Disparity of attainment between different groups of students in UK higher education has been correlated with ethnicity (UUK & NUS, 2019). For example, students who declared their ethnicity as Black were 20% less likely to graduate with a top classification than those who declared their ethnicity as White (OfS, 2018a). The causes of such attainment gaps are complex, and one important factor may be the nature of the feedback given by academic staff on assignments written by different groups of students. This paper aims to explore the feasibility of investigating this hypothesis by analyzing written feedback and looking for patterns in feedback given to different groups of students. Literature Review: Research on attainment among Black and Minority Ethnic (BAME) students in the UK has explored a number of aspects, and has generally concluded that there are issues of “belonging” (Richardson, 2015), particularly in institutions where the majority of academic staff and students are White, but that no single variable can explain the disparity. The wording of feedback on lower-scoring papers has been shown to be more impersonal and distant than that given to students on higher-scoring papers (e.g., Gardner, 2004), which has the (unintended) result of increasing the sense of belonging of higher performing students in ways that can build incrementally over the years of a degree course. While there have been many such small-scale studies of written feedback, none have aimed to collect large quantities of authentic written feedback for analysis. Research Questions: The hypotheses that drive our exploration are that written feedback information (WFI) (Boud & Malloy, 2013) is worded differently to different groups of students, and that there is a direct relationship between this aspect of feedback and academic attainment as measured by grades on summative assessments. Specifically, we asked: 1. Can a framework of WFI functions be developed for our data that share a meaningful set of attributes? 2. Can these categories be used to differentiate WFI to different groups of students? Methodology: A small pilot corpus was compiled from written feedback comments on twelve student assignments from two large Faculties. Metadata was added to each file, and the WFI comments were annotated and analyzed according to a framework developed in a branching format through a recursive construction process informed by the literature reviewed and the data in the corpus. This technique was used to characterize the WFI styles of the two Faculties. Results: The results show that all WFI comments could be classified using the novel systematic framework developed, and that its binary nature enabled ready cross-tabulation with metadata variables. Praise and critique were found to be most frequent, with specific praise of ideas (P1A) accounting for 68% of all praise, and specific critique of content (C1A) accounting for 49% of all critique. Observations tend to be the longest feedback comments (average 15.4 words). When the two Faculties are compared, two different feedback styles are evident, with Fac1 providing more advice, query, and observation style feedback than Fac2, and Fac2 providing more praise and critique than Fac1.

• Background: Researchers have been working towards better understanding differences in professional disciplinary writing (e.g., Ewer & Latorre, 1969; Hu & Cao, 2015; Hyland, 2002; Hyland & Tse, 2007) for decades. Recently, research has taken important steps towards understanding disciplinary variation in student writing. Much of this research is corpus-based and focuses on lexico-grammatical features in student writing as captured in the British Academic Written English (BAWE) corpus and the Michigan Corpus of Upper-level Student Papers (MICUSP). The present study extends this work by analyzing lexical and cohesion differences among disciplines in MICUSP. Critically, we analyze not only linguistic differences in macro-disciplines (science and engineering), but also in micro-disciplines within these macro-disciplines (biology, physics, industrial engineering, and mechanical engineering).\n• Literature Review: Hardy and Römer (2013) used a multidimensional analysis to investigate linguistic differences across four macro-disciplines represented in MICUSP. Durrant (2014, in press) analyzed vocabulary in texts produced by student writers in the BAWE corpus by discipline and level (year) and disciplinary differences in lexical bundles. Ward (2007) examined lexical differences within micro-disciplines of a single discipline.\n• Research Questions: The research questions that guide this study are as follows:\n1. Are there significant lexical and cohesive differences between science and engineering student writing? 2. Are there significant lexical and cohesive differences between micro-disciplines within science and engineering student writing?\n• Research Methodology: To address the research questions, student-produced science and engineering texts from MICUSP were analyzed with regard to lexical sophistication and textual features of cohesion. Specifically, 22 indices of lexical sophistication calculated by the Tool for the Automatic Analysis of Lexical Sophistication (TAALES; Kyle & Crossley, 2015) and 38 cohesion indices calculated by the Tool for the Automatic Analysis of Cohesion (TAACO; Crossley, Kyle, & McNamara, 2016) were used. These features were then compared both across science and engineering texts (addressing Research Question 1) and across micro-disciplines within science and engineering (biology and physics, industrial and mechanical engineering) using discriminate function analyses (DFA).\n• Results: The DFAs revealed significant linguistic differences, not only between student writing in the two macro-disciplines but also between the micro-disciplines. Differences in classification accuracy based on students’ years of study hovered at about 10%. An analysis of accuracies of classification by paper type found they were similar for larger and smaller sample sizes, providing some indication that paper type was not a confounding variable in classification accuracy.\n• Discussion: The findings provide strong support that macro-disciplinary and micro-disciplinary differences exist in student writing in these MICUSP samples and that these differences are likely not related to student level or paper type. These findings have important implications for understanding disciplinary differences. First, they confirm previous research that found the vocabulary used by different macro-disciplines to be “strikingly diverse” (Durrant, 2015), but they also show a remarkable diversity of cohesion features. The findings suggest that the common understanding of the STEM disciplines as “close” bears reconsideration in linguistic terms. Second, the lexical and cohesion differences between micro-disciplines are large enough and consistent enough to suggest that each micro-discipline can be thought of as containing a unique linguistic profile of features. Third, the differences discerned in the NLP analysis are evident at least as early as the final year of undergraduate study, suggesting that students at this level already have a solid understanding of the conventions of the disciplines of which they are aspiring to be members. Moreover, the differences are relatively homogeneous across levels, which confirms findings by Durrant (2015) but, importantly, extends these findings to include cohesion markers.\n• Conclusions: The findings from this study provide evidence that macro-disciplinary and micro-disciplinary differences at the linguistic level exist in student writing, not only in lexical use but also in text cohesion. A number of pedagogical applications of writing analytics are proposed based on the reported findings from TAALES and TAACO. Further studies using different corpora (e.g., BAWE) or purpose assembled corpora are suggested to address limitations in the size and range of text types found within MICUSP. This study also points the way toward studies of disciplinary differences using NLP approaches that capture data which goes beyond the lexical and cohesive features of text, including the use of part-of-speech tags, syntactic parsing, indices related to syntactic complexity and similarity, rhetorical features, or more advanced cohesion metrics (latent semantic analysis, latent Dirichlet allocation, Word2Vec approaches).

Background: A shift of focus has been marked in recent years in the development of automated essay scoring systems (AES) passing from merely assigning a holistic score to an essay to providing constructive feedback over it. Despite all the major advances in the domain, many objections persist concerning their credibility and readiness to replace human scoring in high-stakes writing assessments. The purpose of this study is to shed light on how to build a relatively simple AES system based on five baseline writing features. The study shows that the proposed AES system compares very well with other state-of-the-art systems despite its obvious limitations. Literature Review: In 2012, ASAP (Automated Student Assessment Prize) launched a demonstration to benchmark the performance of state-of-the-art AES systems using eight hand-graded essay datasets originating from state writing assessments. These datasets are still used today to measure the accuracy of new AES systems. Recently, Zupanc and Bosnic (2017) developed and evaluated another state-of-the-art AES system, called SAGE, which enclosed new semantic and consistency features and provided for the first time an automatic semantic feedback. SAGE’s agreement level between machine and human scores for ASAP dataset #8 (the dataset also of interest in this study) was measured and had a quadratic weighted kappa of 0.81, while it ranged for 10 other state-of-the-art systems between 0.60 and 0.73 (Chen et al., 2012; Shermis, 2014). Finally, this section discusses the limitations of AES, which come mainly from its omission to assess higher-order thinking skills that all writing constructs are ultimately designed to assess. Research Questions: The research questions that guide this study are as follows: RQ1: What is the power of the writing analytics tool’s five-variable model (spelling accuracy, grammatical accuracy, semantic similarity, connectivity, lexical diversity) to predict the holistic scores of Grade 10 narrative essays (ASAP dataset #8)? RQ2: What is the agreement level between the computer rater based on the regression model obtained in RQ1 and the human raters who scored the 723 narrative essays written by Grade 10 students (ASAP dataset #8)? Methodology: ASAP dataset #8 was used to train the predictive model of the writing analytics tool introduced in this study. Each essay was graded by two teachers. In case of disagreement between the two raters, the scoring was resolved by a third rater. Basically, essay scores were the weighted sums of four rubric scores. A multiple linear regression analysis was conducted to determine the extent to which a five-variable model (selected from a set of 86 writing features) was effective to predict essay scores. Results: The regression model in this study accounted for 57% of the essay score variability. The correlation (Pearson), the percentage of perfect matches, the percentage of adjacent matches (±2), and the quadratic weighted kappa between the resolved scores and predicted essay scores were 0.76, 10%, 49%, and 0.73, respectively. The results were measured on an integer scale of resolved essay scores between 10-60. Discussion: When measuring the accuracy of an AES system, it is important to take into account several metrics to better understand how predicted essay scores are distributed along the distribution of human scores. Using average ranking over correlation, exact/adjacent agreement, quadratic weighted kappa, and distributional characteristics such as standard deviation and mean, this study’s regression model ranks 4th out of 10 AES systems. Despite its relatively good rank, the predictions of the proposed AES system remain imprecise and do not even look optimal to identify poor-quality essays (binary condition) smaller than or equal to a 65% threshold (71% precision and 92% recall). Conclusions: This study sheds light on the implementation process and the evaluation of a new simple AES system comparable to the state of the art and reveals that the generally obscure state-of-the-art AES system is most likely concerned only with shallow assessment of text production features. Consequently, the authors advocate greater transparency in the development and publication of AES systems. In addition, the relationship between the explanation of essay score variability and the inter-rater agreement level should be further investigated to better represent the changes in terms of level of agreement when a new variable is added to a regression model. This study should also be replicated at a larger scale in several different writing settings for more robust results.

Background: This study focuses on construct representation and inter-reader agreement and reliability in ePortfolio assessment of 1,315 writing portfolios. These portfolios were submitted by undergraduates enrolled in required writing seminars at the University of Pennsylvania (Penn) in the fall of 2014.  Penn is an Ivy League university with a diverse student population, half of whom identify as students of color. Over half of Penn’s students are women, 12% are international, and 12% are first-generation college students. The students’ portfolios are scored by the instructor and an outside reader drawn from a writing-in-the-disciplines faculty who represent 24 disciplines. The portfolios are the product of a shared curriculum that uses formative assessment and a program-wide multiple-trait rubric. The study contributes to scholarship on the inter-reader reliability and validity of multiple-trait portfolio assessments as well as to recent discussions about reconceptualizing evidence in ePortfolio assessment.  Research Questions: Four questions guided our study: What levels of interrater agreement and reliability can be achieved when assessing complex writing performances that a) contain several different documents to be assessed; b) use a construct-based, multi-trait rubric; c) are designed for formative assessment rather than testing; and d) are rated by a multidisciplinary writing faculty?   What can be learned from assessing agreement and reliability of individual traits? How might these measurements contribute to curriculum design, teacher development, and student learning? How might these findings contribute to research on fairness, reliability, and validity; rubrics; and multidisciplinary writing assessment? Literature Review: There is a long history of empirical work exploring the reliability of scoring highly controlled timed writings, particularly by test measurement specialists. However, until quite recently, there have been few instances of applying empirical assessment techniques to writing portfolios.  Developed by writing theorists, writing portfolios contain multiple documents and genres and are produced and assessed under conditions significantly different from those of timed essay measurement. Interrater reliability can be affected by the different approaches to reading texts depending on the background, training, and goals of the rater. While a few writing theorists question the use of rubrics, most quantitatively based scholarship points to their effectiveness for portfolio assessment and calls into question the meaningfulness of single score holistic grading, whether impressionistic or rubric-based. Increasing attention is being paid to multi-trait rubrics, including, in the field of writing portfolio assessment, the use of robust writing constructs based on psychometrics alongside the more conventional cognitive traits assessed in writing studies, and rubrics that can identify areas of opportunity as well as unfairness in relation to the background of the student or the assessor. Scholars in the emergent field of empirical portfolio assessment in writing advocate the use of reliability as a means to identify fairness and validity and to create great opportunities for portfolios to advance student learning and professional development of faculty.  They also note that while the writing assessment community has paid attention to the work of test measurement practitioners, the reverse has not been the case, and that conversations and collaborations between the two communities are long overdue. Methodology: We used two methods of calculating interrater agreement: absolute and adjacent percentages, and Cohen’s Unweighted Kappa, which calculates the extent to which interrater agreement is an effect of chance or expected outcome. For interrater reliability, we used the Pearson product-moment correlation coefficient. We used SPSS to produce all of the calculations in this study.  Results: Interrater agreement and reliability rates of portfolio scores landed in the medium range of statistical significance.  Combined absolute and adjacent percentages of interrater reliability were above the 90% range recommended; however, absolute agreement was below the 70% ideal.  Furthermore, Cohen’s Unweighted Kappa rates were statistically significant but very low, which may be due to “kappa paradox.” Discussion: The study suggests that a formative, rubric-based approach to ePortfolio assessment that uses disciplinarily diverse raters can achieve medium-level rates of interrater agreement and reliability. It raises the question of the extent to which absolute agreement is a desirable or even relevant goal for authentic feedback processes of a complex set of documents, and in which the aim is to advance student learning. At the same time, our findings point to how agreement and reliability measures can significantly contribute to our assessment process, teacher training, and curriculum. Finally, the study highlights potential concerns about construct validity and rater training.  Conclusion: This study contributes to the emergent field of empirical writing portfolio assessment that calls into question the prevailing standard of reliability built upon timed essay measurement rather than the measurement, conditions, and objectives of complex writing performances.  It also contributes to recent research on multi-trait and discipline-based portfolio assessment.  We point to several directions for further research:  conducting “talk aloud” and recorded sessions with raters to obtain qualitative data on areas of disagreement; expanding the number of constructs assessed; increasing the range and granularity of the numeric scoring scale; and investigating traits that are receiving low interrater reliability scores. We also ask whether absolute agreement might be more useful for writing portfolio assessment than reliability and point to the potential “kappa paradox,” borrowed from the field of medicine, which examines interrater reliability in assessment of rare cases. Kappa paradox might be useful in assessing types of portfolios that are less frequently encountered by faculty readers. These, combined with the identification of jagged profiles and student demographics, hold considerable potential for rethinking how to work with and assess students from a range of backgrounds, preparation, and abilities.  Finally, our findings contribute to a growing effort to understand the role of rater background, particularly disciplinarity, in shaping writing assessment. The goals of our assessment process are to ensure that we are measuring what we intend to measure, specifically those things that students have an equal chance at achieving and that advance student learning.  Our findings suggest that interrater agreement and reliability measures, if thoughtfully approached, will contribute significantly to each of these goals.