Hello from the “Other Side”!
Designing UX for Conversational Recommendation Systems
Size does matter
Day by day, enterprises populate their catalog with a variety of solutions and products. Users often find themselves scrolling through a plethora of products, card by card, section by section, only to be lost in filters, sorts, and comparisons. Such a cognitive load leads to increased bounce rates.
Enter Recommendation Systems
With advancements in AI(Artificial Intelligence), NLP(Natural Language Processing) & ML(Machine Learning), enterprises have begun introducing chat-bots/assistants which act as recommendation systems. They help narrow down the choices for the users/customer support teams thus streamlining the filtering process.
The Human-Machine Symbiosis
Wait! No, not Cyborg.
It’s the HCI these days. It is not just the end user who benefits from the recommendation systems. Inputs from the users act as reinforcements to the internal decision making of these engines and that’s why they tend to produce better results in the long run. Feedbacks are a crucial subset of the inputs that can affect the performance of an AI/ML engine. Ironically, users are often ignorant of the role of feedbacks and thus are reluctant to give the same. The engines would improvise only so much without the feedback. Thus, we need machines which in turn need our feedback, in this symbiosis.
The engines would improvise only so much without the feedback.
Getting the user feedback without disrupting the workflow is crucial: It has to happen on-the-fly and within that bare minimum window. Remember, at this stage feedback is probably the last thing on the user’s mind. All they’re looking for is the right product options. Instead, the user is more likely to give a feedback when they can empathise with the machine. The interactions must be designed in such a way that they trigger the feedback system at just the right time and the same should be supported with micro-texts.
The user is more likely to give a feedback when they can empathise with the machine.
Good Things Take Time
Its not about 2nd chances or 3rd chances or even 1000th chances. Its about improving gradually with every input. Just how Elon’s OpenAI defeated the DOTA2 world champion.
The trick wasn’t to teach the AI how to play the game — instead, it was to have the bot play many games against itself, and encourage ideal behaviour as it learned the ropes over time.
So its well established that feedbacks and iterations help shape up such products: It is very important to have human overrides/corrections in order to expect better outputs the next time.
Do checkout the app video underneath and stay tuned for some more interesting content STRAIGHT OUTTA F1
Thanks for your time!
