Josh Feast, is the CEO and Co-Founder of Cogito, an enterprise that combines Emotion and Conversation AI into an innovative platform that provides real-time coaching and guidance to contact center agents, gives supervisors visibility into live conversations from their teams working from anywhere, and continuously monitors customer and employee experiences.
The story of Cogito starts in 1999, before the company was founded. Could you share some insights on these early days at the MIT Human Dynamics Lab and what was being worked on?
From 1999 through 2006, Dr. Sandy Pentland developed fundamental basic science demonstrating the presence and power of social signals in human communication and the ability of machines to detect and interpret them.
In 2007, Cogito was spun out from the MIT Media Lab. Could you share this genesis story?
Before my days at MIT, I recognized the need for technology that is informed by conversational context to aid its users throughout emotionally charged situations. While working at the New Zealand Department of Child, Youth and Family Services (now known as the Child, Youth and Family unit of the Ministry of Social Development), I noticed that many social workers were burnt out due to the highly emotional nature of their duties and believed that the management systems that supported them would greatly benefit from such a technology. I brought my observations from that time to MIT, and Cogito was later created from Dr. Pentland’s MIT Media Lab research that seemed to directly address the problem. Cogito received funding from the Defense Advanced Research Projects Agency (DARPA) to research and develop an artificial intelligence platform and behavioral models to automatically detect human psychological states. This technology proved successful at helping military veterans returning from combat through deployments with the Department of Veteran Affairs (VA).
The Emotion AI technology that is used at Cogito was first validated by assisting healthcare providers to detect early signs of PTSD and other mental health disorders in soldiers returning from combat. Could you discuss some details regarding this and the types of results that were seen?
The aim of deploying this technology to healthcare providers was to detect depression and prevent suicide in military veterans returning from combat. The platform we developed enabled doctors to track veterans’ overall mental health through voice signals and to pinpoint events like homelessness and other warning signs of poor mental health. We quickly realized we had something special, and that the technology’s application could prove useful beyond supporting military veterans and healthcare systems in areas with high volumes of complex, emotionally charged conversations. With our roots still centered on the human experience, we became the Cogito you know today, supporting real-time coaching and guidance for large scale enterprise contact center agents across multiple industries including healthcare.
Can you discuss how Cogito uses AI to analyze behavioral cues and provide in-the-moment feedback during conversations?
Cogito uses a powerful combination of Emotion and Conversation AI to reveal new insights from all conversations, extracting both what was said and how the customers received the message. These AI models measure customer experience (CX) in real-time on all calls to have impacts in the moment, vs. post-call analysis which centers only on improving future interactions.
Cogito extracts and analyzes more than 200 acoustic and voice signals in milliseconds to give contact center agents cues on how to adjust their behavior and surface the best recommendations based on the topics discussed and desired outcomes.
Cogito performs live, in-call voice analysis to augment behavior in real-time to create better human connections at scale between customers and contact center agents, regardless of where they work.
How does this feedback guide agents to build better relationships with customers?
The real-time feedback contact center agents receive from Cogito’s nudges allows agents to display more consistent emotional intelligence, which results in agents delivering empathy on each call. Improved empathy leads to better conversational outcomes, such as reduced call handle times, increased first call resolution, improved customer satisfaction, and increased customer lifetime value.
Each contact center representative has different strengths and weaknesses. The real-time nudges they receive on the call helps enhance their customer service, whether it be to provide more empathy, speak slower, or sound more upbeat. This tailored feedback in the moment allows agents to build a relationship with the customer based on that particular customer’s experience and their voice signals picked up by the AI model. In turn, this improves both the customer experience, and the agent experience.
Real-time feedback is not only beneficial to CX – it also benefits the employee experience (EX). Our tools help support representatives to have more positive work experiences, which is proven to drive higher levels of CX.
In 2019, Cogito released a paper titled “Gender de-biasing in speech emotion recognition.” What were some of the key insights when it came to the effect of gender bias in speech with respect to emotion?
Our paper focused on the modeling approach and optimization techniques as well as sampling bias. Therefore, more research must be done to mitigate negative bias generally in machine learning and specifically in speech emotion recognition. Key insights include:
Female speech tends to be higher pitch than male speech, which results in more widely spaced harmonics.
Speech emotion recognition models can be affected by this difference. This can lead to lower accuracy for female speech versus male speech.
De-biasing machine learning techniques can be applied to reduce this accuracy imbalance. In the paper, Cogito introduces a novel de-biasing technique which performs favorably relative to the baseline.
How does Cogito operate to mitigate the effects of unwanted gender or other types of bias?
Cogito uses natural language processing (NLP) models that combine human-aware AI systems, deep learning machine models, and other complex rules which help computers understand, analyze, and simulate human language. We are consistently working on and evolving our NLPs with new data to mitigate bias.
Cogito has a comprehensive protocol for machine learning model development, which aims explicitly at mitigating bias and ensuring ethical machine learning (ML)-based product features. This protocol covers areas like sampling data for training, mitigating bias in human labeling, and using ML de-biasing techniques.
Cogito uses a ‘fairness’ dataset comprised of a large body of audio data where the speakers self-report different demographic categories. All models are assessed against the fairness dataset and against the various demographic categories. We also use ML Ops techniques to objectively monitor models in production and systematically carry out model audits with human annotation.
What are your personal views on how AI shouldn’t only replace humans, but rather augment human behavior?
There are things humans can do and nuances they can provide in human-to-human interactions that technology like AI cannot emulate on its own. For example, customers want to receive empathy when they contact customer support. If the customer interacts only with an automated system powered by AI, their issue might be resolved, but they could end up feeling frustrated or annoyed by the interaction. If we replace all contact center agents with AI, then we are eliminating the human element that’s necessary to build relationships and achieve and maintain lasting, loyal customers.
When engaging in a service interaction, humans value talking to someone who can put themselves in their shoes, who has had experiences similar to what they themselves are going through. Along the same lines, humans value the sensation of someone else taking care of them and owning the resolution to their problem. It will be a long time before standalone AI is really perceived as something other than a self-help tool.
Is there anything else that you would like to share about Cogito?
At Cogito, we’re developing new technologies to usher in the next generation of contact centers. Earlier this year, we released our Employee Experience (EX) Score to track agents’ experiences. Similar to our customer experience (CX) score, the EX Score combines human-aware Emotion AI and Conversation AI, deriving real-time insights across single instances or trends across multiple calls. Amid high levels of dissatisfaction, burnout, and attrition, the EX Score helps solve the question of how to prevent burnout and help the agent experience, which in turn drives better customer experiences and long-term business sustainability.
Thank you for the great interview, readers who wish to learn more should visit Cogito.
- DeepFace for Advanced Facial Recognition
- Mohan Giridharadas, Founder & CEO at LeanTaaS – Interview Series
- ContentStudio Review: The Best AI Tool for Social Media? (September 2023)
- Decoding Opportunities and Challenges for LLM Agents in Generative AI
- Tailor Introduces ChatGPT Plugin Enabling Conversational Interface for ERP Operations