Conversational AI to Accelerate Skill Integration

Conversational AI to Accelerate Skill Integration
Project Type: 
Academic Technology
App Development
Tech Innovation Initiative
Natural Language Processing
Language acquisition
Guadalupe Valdes
5 months
Josh Weiss

Language coaches converse with chatbots as part of a training curriculum, in the process iterating their knowledge and evaluating their gaps



Guadalupe Valdes, professor at the GSE, identified a gap in language coaches’ capacity to identify proficiency levels in adult learners. Despite a rubric and instructions, coaches found it difficult to assess the language proficiency of another adult when faced with a real person, frequently citing lack of practice. 


Conversational AI was selected as a solution based on its ability to simulate language patterns and adapt to real-time language input. In effect, a chatbot would be a practice partner to practice interactions and language assessment scenarios with non-native adult English speakers.

DialogFlow, a chatbot engine, was selected for its ease-of-use, machine learning capabilities, and rapid integration of user data. The schema for each chatbot was ideated with Dr. Valdes in order to reflect the instructional purpose. Prototypes were developed and beta tested in order to hone domain scope, variation in response types, formality of language, and logic branching.

Technologies used 

Google Cloud Platform


The chatbots were rolled out in the Ingles Personal curriculum starting in the Summer 2019 term. A GSE offering, Ingles Personal is a hybrid course -- partly online, partly in-person -- with rich interactive elements to promote active learning. The chatbots were integrated into the interface of the course, and were bookended by an introduction to language proficiency diagnosis as well as a cohort-wide reflection on the chatbot experience. 

Chatbot rubric
The three chatbots were modeled on a language proficiency rubric designed by Dr. Valdes.
Stanford volunteers and adult learners pair up to improve English proficiency. The program recently underwent significant changes to better train language coaches.
Each chatbot simulates a different level of English proficiency. Language coaches role play with each chatbot to practice their evaluation skills.