Advancement in cognitive technologies is causing us to rethink and restructure every experience we build as a dynamic model based on machine augmented user interaction.  The success of machine learning is highly dependent on user engagement which in its turn relies on seamless cognitive augmentation that fulfills user goals.  Understanding both sides of this process during design and development stages leads to products with AI grounded in human needs and solving for them in ways uniquely possible through machine learning.

I was tasked with directing UX design for Noetos, the leading open source machine learning and artificial intelligence platform that assists data scientists i building interpretable models and provides shareable insights for enterprise intelligence to establish and sustain  positive feedback loop known as “more data >> better AI >> more mass adoption, repeat.”  

My approach was agile and based on design sprints to further extend positive impact of digital loops making them adaptive to users’ needs and directing collaborative ideation to achieve AI/UX alignment.

If AI and UX are not properly aligned cognitive augmentation produces distortion that results in misuse and frustration.   AI by itself cannot determine which problems to solve.  If it is not aligned with a human need, a resulting powerful system addresses marginal issues.  Users often develop mental models that suit their imaginary theories about AI resulting in misuse and lack of trust.  This phenomena is usually caused by design flaws related to AI targets being underdefined causing lack of understanding in user role of calibrating that system.

Qualitative Elevation

In order to thrive, the platform needed to acquire multi-dimensional approach that includes social and technical perspectives. Machine learning is the science of making predictions based on patterns and relationships that’ve been automatically discovered in data.  From model development, to the source of data, samples, descriptors all the way to success criteria every facet of AI is affected by human judgement.  And approaching AI from human-centered perspective allows qualitative elevation that for businesses translates to quantifiable returns.

Building for Real Humans: Augmenting Analysts Performance

AI aligned with UX translates in powerful cognitive augmentation and addresses a real human need the way humans need it addressed.  Our team defines this alignment as “let people do what they do best and let machines do what people do worst … because in order for us to build trust in the impact of AI we must feel reassured, included and informed.”

Cognitive Load Mapping
During the design sessions, I mapped customer journeys onto UI sections. Participants spent 10-15 minutes writing down all known cognitive spike points and assigning load levels.

After each person generated spike points, I facilitated a group discussion and an activity where we collectively mapped the pain points onto a visual template representing UI touch points of the customer flow. The higher or more severe the pain point, the higher up we would place the circle. This process of assessing the impact of pain points helped us in determining which areas to focus on as we began to define the scope of our human-centered AI transformation.

Abstracting Experience Layer: Decoupling ML and UX Optimization
I made a decision to prototype our UX with prebuilt models rather than testing with real ML models. The latter takes an incredibly long time to build and instrument (and is far less agile or adaptive than traditional software development, so it’s more costly to swing and miss), while the former afforded us genuine insights into the way people will derive value and utility from our (theoretical) product.

Categorizing Cognitive Loads and UI Mapping and Compensation
The above graphs illustrate that while most areas have at least some learning curve, with the added overhead of AI, it’s especially important to ‘spend’ wisely on your user’s cognitive load.  When the context of use is novel to the user [figure A], bias for dependability is warranted.  When there are a lot of new UI elements to learn [figure B], accentuating familiarity of use cases is needed.  Dynamic functionality of the product  [figure C] calls for reinforcing familiar patterns in UI.

Cleaning up Development Bias
While going through the process of cognitive optimization I noticed that our UX was biased for increasing complexity in correlation with deeper development of the analytics platform.  The complexity of an interaction model accelerated with the complexity of the system it’s driving. But that’s sort of where we were for awhile during our early design phase, and we got away with it in large part for three reasons:

• We exposed people to simulated models in a usability lab.
• Surrounded by the developers and designers we were losing touch with the reference points that everyone else would bring to the table.

Cognitive distribution map

To compensate for development bias I began fiercely reducing complexity in the UI, and made control and familiarity cornerstones of our experiential framework.  I made sure that the user had the final say in curation; from allocation to portfolio management. I also suggested optimizing for recall and we showed analysts more options than what we necessarily thought was just right, because by allowing them to look a bit below the ‘water line’ and reject stuff they didn’t want, they actually developed a better understanding of what the system was looking for, as well as what they could confidently expect it to capture in the future.

Test Driving Hybrid Enterprise Intelligence

Cognitive performance before optimization

Cognitive performance after optimization

By re-orienting the conventional AI  paradigm from finding ways to make the machine smarter, to exploring ways to augment human capability, we unlocked far greater potential in machine learning augmenting analysts performance in the process, increasing value and elevating customer experience.