Ethical Shopping Recommender

• 

  High level design

• 

  Workflow illustration (from Twilio UI)

Inspiration

I am a researcher in RRI (responsibility research and innovation) and was motivated by Capital One's challenge to use banking data for good. I also came to develop my (weak!!) programming skills and experience a Hackathon.

What it does

*Identifies alternative options based on shopping transaction history using open-web Corporate Social Responsibility company ratings (from CSRHub) *Provides a link to detailed responsibility assessment for a company, and alternatives listed by responsibility rating. *Makes one recommendation at a time so as not to flood the customer *Allows opt out of the recommender service. *Links the caller to Customer service centre if they want to be connected *Error capture in case any of the services fail Original plan was also to link the ratings to other services e.g. Trustpilot, Glassdoor and Twitter sentiment but descoped due to lack of time and lack of access to commercial APIs other than Twitter

How I built it

Twilio workflow API connection to Capital One Google sheets library of companies to CSRhub ratings (not used)

Challenges I ran into

Capital One API doesn't provide Merchant ID attribute which would be necessary No access to CSRHub API. Would have needed to webscrape the ratings Initially tried a workflow using MS Flow to Twilio but couldn't track the call state over multiple back and forth HTTP calls Couldn't see how to pick up elements from prior workflow stages in Twilio Couldn't successfully call the API from Twilio (unclear whether this is due to a Twilio side issue). High latency on SMS Twilio responses in testing (possibly due to use throttling) No working screen capture app appears to be available following IOS13 update (tried five)

Accomplishments that I'm proud of

Developed an SMS chatbot workflow (json is in the github repo) Developed a library of company name-website-twitter handle-CSRhub page Downscoped initially overambitious proposal (albeit not enough)

What I learned

What a hackathon is like How to use Twilio First use of github Developed more experience with API use and testing

What's next for Ethical Shopping Recommender

*Recommend Capital One see if a Merchant ID attribute is available and can be added to their dev API *If product level data is available (e.g. SKU) this would allow for very granular recommendations *If customer attributes are made available, a full AI recommender system could be implemented which learns from the customer's own preferences and those of similar customers *If customer had the choice, they could either opt in to recommends for particular product/service types, or to indicate their priorities (e.g. environmental, equality/diversity) which could be used to upweight their recommendations

Built With

• capitaloneapi
• csrhub
• twilio