Document Type

Conference Proceeding


LAK19: The 9th International Learning Analytics & Knowledge Conference

Publication Version

Published Version

Publication Date


First Page


Conference Title

Companion Proceedings 9 th International Conference on Learning Analytics & Knowledge (LAK19)

Conference Date

March 2019


Tempe, AZ


Despite the appeal of automated writing evaluation (AWE) tools, many writing scholars and teachers have disagreed with the way such tools represent writing as a construct. This talk will address two important objections – that AWE heavily subordinates rhetorical aspects of writing, and that the models used to automatically analyze student texts are not interpretable for the stakeholders vested in the teaching and learning of writing. The purpose is to promote a discussion of how to advance research methods in order to optimize and make more transparent writing analytics for automated rhetorical feedback. AWE models will likely never be capable of truly understanding texts; however, important rhetorical traits of writing can be automatically detected (Cotos & Pendar, 2016). To date, AWE performance has been evaluated in purely quantitative ways that are not meaningful to the writing community. Therefore, it is important to complement quantitative measures with approaches stemming from a humanistic inquiry that would dissect the actual computational model output in order to shed light on the reasons why the ‘black box’ may yield unsatisfactory results.


This proceeding was published as Cotos, E. Understanding the ‘Black-Box’ of Automated Analysis of Communicative Goals and Rhetorical Strategies in Academic Discourse., LAK19: Proceedings of the 9th International Conference on Learning Analytics & Knowledge.,Tempe, AZ., March 2019.

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.

Copyright Owner

The Author



File Format



Article Location