I recently attended a JISC event @ Bolton College which discussed the usage of chatbots at the institution (thank to Jisc and Bolton for the hospitality).

Bolton are doing  alot with chatbots and are using IBM Watson Conversation (which is now rebranded as Watson Assistant)  and starting to expand from text to voice platforms (e.g. Alexa).

It;s a great example of how bots can be used to provide users with instant responses to general questions. Its the creation of an automated support layer that replaces the need to click through numerous webpages or flick through a course handbook (if a student can remember where it is on top of everything else!). It also replicates what some students experiences and expect of service providers (e.g. ASOS, Tesco, Skyscanner all use bots for support, booking and general queries).

However I have concerns for topic based support and how we envisage the usage of bots\automation for teaching students. Is learning solely meant to be the regurgitation of content that meets the initial question or is it being able to critique and apply knowledge and question what we know?

Plus with the ongoing commercialisation of education and the competitve marketplace of global, private and public institutions could we start to see the introduction of bots to faciliate quicker & cheaper teaching?  Could this lead to the loss of knowledge expansion with automation accepting the norm and rejecting new concepts or theories? (hence the quote below), or could we see the reduction in quality in teaching with automated qualifications that have no human interaction and can be delivered at scale and on a cheaper budget (and what would be the quality or worth of these qualifications)? And i want start on ethical concerns (this is meant to be a short post, not the bible).

A computer program might be able to survey the literature and figure out which questions remain, but would it be able to pick out research of Einsteinian proportions—some new theory that completely upends previous assumptions about how the world works? – https://www.wired.com/2017/02/ai-can-solve-peer-review-ai-can-solve-anything/

Anyway back to Bolton, my notes are below, my thoughts are for another post:

• Use IBM Watson for established student bot and in development staff bot (called Ada)
• IBM Watson is nice and easy to use and again like the other big 4 it wouldn’t take long to build (however the conversation mapping is the leg work which would be ongoing).
• In use since 2017
• Branded as a virtual assistant
• Embedded on webpages with voice option
• 800-900 enquiries per day to Ada (at the start of the term)
• Bespoke ‘Ask Ada’ apps for IOS and android to be released 2019 (emulate desktop version)
• Ada has been developed with a personality and gender
• Alexa skill in development
• Bolton see the rise of bots disrupting Learning tech and questioned the relevance of a LMS in 5 years
• They have embedded Ada into Moodle
• Queries split into the following:
  • General: about the college
  • Specific: about the user (requires auth//progress monitoring//what is my email address)
  • Topic: topic based information (videos and or definitions of subjects for 3 courses currently however expanding)
• Queries routed via IBM answered by the following (and in the priority):
  • IBM Watson api for configured Q&As (e.g. intents\entities)
  • Live Internal data set (for specific queries)
  • https://www.wolframalpha.com/ (for all other queries, it’s a Q&A site)
• Model used below:
  • Bolton stated no personal data is shared with IBM as its all hashed with placeholders used (see image below, excuse my scribbles)

• Linked to Timetabling which can pull coming classes and attendance
• Traffic light system for attendance (Q: How am I doing? A: traffic light per course showing attendance record)
• Queries are Q&A based, couldnt see any follow-up from the bot (e.g. your attendance is displayed as red traffic light, nothing is prompted from the bot regarding support etc)
• Aftab (head of IS team) spoke a lot about calm technology and the symbiotic relationship it should have with users
  • See voice taking over and pushing more voice options (#voicefirst)
• Developing a community of bots that specialises in areas (admin tasks, teaching, support)
• Assessment bot being developed
  • Initially for work experience evaluation where students need to provide a short example of meeting objectives (collaboration, problem solving via a para)
  • Answers uploaded via webpage and score and feedback returned instantly
  • Using NLC (Natural learning classifier aka what is being expressed in text and subsequent categorisation into classes) & NLU with configured feedback\score based on the related action (from the derived classification)
  • Possible pre-submisssion tool to gauge score was discussed (could this lead to the gamification of score)
  • Slide for process below:

Nov 15, 2018

Share