Can Alexa Predict If Your Relationship Will Last?
Advances in AI may reveal insights into interpersonal relationship functioning.
Posted Feb 22, 2019
Do you have an Amazon Echo or a Google Home device in your home? If so, you're probably comfortable having your conversations — at least to some extent — monitored by technology. This includes conversations you have with your romantic partner, from words of love and witty banter to strained exchanges and full on arguments. The tone you both use, the pauses between words, the volume of your voice — all of these variations could be extracted and recorded as data with the right technology. Could a synthesis of these data give you clues to your relationship's health?
Technology's Potential for Monitoring Relationship Well-Being
Relationships can come with quite a bit of uncertainty, particularly if you've hit a rough patch, or if you're trying to determine if your partner is the person for you. Maybe a device that monitors your interactions could help you figure this out.
The idea that the technological monitoring of your intimate conversations could give insight into your relationship's health is compelling and not so far-fetched. Think about how we use words differently depending on how we feel about the person we're speaking to. "Where were you?" can be a kind, benign inquiry or an aggressive, accusatory question, depending on how we ask it. The person you're speaking to will likely pick up on your pitch, tone, and volume, and so too might a sufficiently sophisticated technology.
Maybe there are patterns that happy couples have which unhappy couples don't, or the reverse. In that case, maybe technology can help you predict whether you're likely to stay with your partner or break up.
The Potential for AI to Offer Relationship Insight
Research in artificial intelligence (AI) suggests that your conversations could include acoustical patterns that shed light on the quality of your romantic relationship, and that computer algorithms can read and interpret these patterns. One study analyzed three sessions of video-recorded data collected over a two-year period from 134 married couples in distressed relationships while they discussed a critical issue in their relationship (Nasir et al., 2017). These recordings were part of a clinical study, and couples knew they were being recorded. The couples also completed a variety of marital outcome assessments.
Of interest: Can an algorithm be programmed to predict whether couples will stay together based on their vocal speech pattern? And, if so, could the algorithm do this better than human experts? To test the latter question, Nasir and colleagues (2017) trained experts to identify and code videos of the couples for clinically relevant behavioral patterns (e.g., blame patterns, negativity).
So, what did they find? Interestingly, the algorithm predicted relationship outcomes better than chance (with 79 percent accuracy). Further, even though the human coders had access to the full video (e.g., vocal patterns, non-verbal behaviors, the content of the conversation, physical movement), and the algorithm relied only on vocal patterns, the algorithm very slightly out-performed the clinical experts at predicting relationship stability. At a minimum, the technology works as well as an expert therapist.
What Does This Mean for Couples?
Down the road, it might be possible to enable a "relationship assessment" skill to our smart devices. In some ways, having a computer "read" our relationships is not much different from having friends or family observe our interactions with our partner. Indeed, friends can often predict whether their friends' relationships will break up better than the people in those relationships (Agnew, Loving, & Drigotas, 2001).
And it might help us, too. We might be biased when we think about our relationship and its potential, and we could be highly motivated to ignore negative signs, or we could over-emphasize the big fight we had last week. We also physically cannot integrate huge volumes of data the way computer algorithms can. A technology that could take "big data" and see it for the patterns it offers could be in a position to tell us how likely it is for our relationship to succeed. Whether we would want that insight, however, is a different question.
Nasir, M., Baucom, B. R., Georgiou, P., & Narayanan, S. (2017). Predicting couple therapy outcomes based on speech acoustic features. PloS one, 12(9), e0185123.
Agnew, C. R., Loving, T. J., & Drigotas, S. M. (2001). Substituting the forest for the trees: Social networks and the prediction of romantic relationship state and fate. Journal of personality and social psychology, 81, 1042-1057.