Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

Using AI to Predict the Health of a Relationship

DZone's Guide to

Using AI to Predict the Health of a Relationship

It's unlikely that we'll be seeing purely AI-based therapy anytime soon, but this research does provide an interesting glimpse into the capabilities of AI systems.

· AI Zone
Free Resource

Find out how AI-Fueled APIs from Neura can make interesting products more exciting and engaging. 

Last year, I wrote about a fascinating project that was unleashing AI in therapy sessions. The software was put through its paces across over 1,000 therapy sessions, with the machine learning algorithm capable of producing an empathy score for each session.

Central to this was a deepening understanding of language. It's been followed by a second paper that was published recently that also used an automated system to perform speech analysis.

The researchers analyzed recorded therapy sessions over a two year period for things such as pitch, intonation, and variation in pitch. They chose these features of speech as they provide strong insight into things such as intensity and tone of speech.

They then trained the algorithm to be able to detect the connection between these vocal features and the outcome of the therapy for the couples. Suffice to say, this was a challenging undertaking and required not just an analysis of the voices but also the interplay of the conversation and things such as who spoke when, and indeed for how long. The machine was not monitoring the actual words at all, but only how they were spoken.

AI Therapy

Suffice to say, that was not actually needed, as the AI system was able to predict with high levels of accuracy whether the couple would stay together purely from the patterns of speaking.

Where things get interesting is that the video recordings were also given to a panel of human experts to provide their own judgments on what they saw and heard. Suffice to say, their method of assessment was somewhat more involved than the AI, using a psychological assessment based upon things like body language, the actual words spoken, as well as the patterns of speech.

Who did better? Well, it was pretty close, with the human experts getting the outcome right 75.6% of the time, but the AI system got the prediction right 79.3% of the time. As with many of these AI systems however, the best results came when the AI was working in partnership with the experts. In that scenario, the pairing was correct 79.6% of the time.

It's another example of not only how effective AI can be at understanding language, but also how valuable such systems can be in detecting patterns that are missed by human experts. It also highlights how much meaning we communicate without realizing it.

Suffice to say, it's unlikely that we'll be seeing automated, or even augmented, therapy anytime soon, but it does nonetheless provide an interesting glimpse into the capabilities of AI systems, and a possible direction of travel.

To find out how AI-Fueled APIs can increase engagement and retention, download Six Ways to Boost Engagement for Your IoT Device or App with AI today.

Topics:
ai ,machine learning ,algorithm ,automation ,speech analysis ,predictive analytics

Published at DZone with permission of Adi Gaskell, DZone MVB. See the original article here.

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}