Can You Change? Measuring Attachment and Hope for the Future

Professor Gillath: That’s the million-dollar question. Our assumption was that, similar to language acquisition, there’s a critical period that shapes who you are and what you are. I mentioned Amir Levine – we just had this conversation last week: how much your childhood is shaping and locking in your attachment.

If you look at the research, the effect sizes are not huge. Actually, they’re pretty small – like many other things in psychology. The connection between childhood attachment and adult attachment is not super strong. A lot of the work by people like Chris Fraley, Bill Chopik, Jeff Simpson and others is tying attachment across the lifespan. Continuity from childhood to adulthood is imperfect and varies with the measures and the time interval. Early relationships matter, but later experiences matter too.

For years, the assumption was that attachment style is being shaped early on. Early experiences were assumed to add a lot, because the child has very little experience or knowledge to begin with or to compare. So the conception in the field was that early interactions in childhood are going to have a large enduring impact. But recent research shows that as people grow older, get into new relationships, new setups, those contribute as well. Bowlby said attachment is ever-changing. It’s a dynamical system. We show these changes in our studies. We see it in developmental work.

Things can change.

Therapy – obviously therapists believe the big thing in therapy is the ability to change. If you can’t change, then why bother with therapy? But this is something we’re still trying to figure out. How much of attachment is set, and how much of it is changing? A lot of the work I do looks at state attachment – the fluctuations in attachment over time – and how much movement you need before an enduring change occurs.

One of the constructs people talk about in the attachment domain, especially from a clinical perspective, is earned security. What happens if you grew up insecure, and then went through therapy or a secure supportive relationship? Do you earn a new sense of attachment security? Is that the same as growing up secure? The evidence to support earned security is complex – causing some scholars to say that all the models (secure, anxious, avoidant) were there all along, but the insecure model was more salient and then secure became the salient one.

We don’t fully know how much of one’s attachment style is based on childhood and how much is changing over time. But in my research with Chris Fraley, and Pascal Deboeck¹ based on the NSF grant we secured we examined change in people’s attachment over time – we found that 50% of the events that we studied result in a change to people’s attachment. And about a quarter of these events make enduring changes – things that stay with us. A very painful breakup. Or being betrayed by your partner can be traumatic and have a long-term effect toward insecurity. Whereas something really nice that your partner did for you, or meeting long-time friends are likely to change it toward security.

So there are events that are very meaningful and have an impact and change things, and then there are events that are just a little blip on the face of your lifespan, and then you move on and forget about them.

Professor Gillath: We have some research on this with Bill Chopik and Sara Konrath⁵. They have a paper that looked from 1988 to about 2011. We’re currently extending that project, looking at whether security goes up or down, whether anxiety goes up or down etc.

The original paper showed that attachment security goes down, whereas avoidance goes up and our current project seems to support these trends. The question is why. People like Jonathan Haidt talk about screens and social media as the culprits. They might be contributing, but the trajectory seemed to have started before the rise of screens and social media into our lives.

Is that fear of relationships? It’s not exactly the same. But attachment anxiety is the fear about being rejected, and abandoned, and not being loved. So there’s obviously an overlap there. This fits with the overall tendency toward social disconnect we mentioned earlier.

Professor Gillath: Measures are fascinating to me. There are so many measures of attachment out there, and there’s a divide between developmentalists and social psychologists on which ones to use. For example, the AAI (the Adult Attachment Interview) and the ECR² (experiences in close relationships) are the flagships of developmental- and social psychological research on attachment respectively. They come from different research traditions, and studies using both often find only weak associations⁶. When you put the two in the same study, they give you very different results – often their scores are not even correlated. Furthermore, if you look at brain activation associated with the ECR versus the AAI, you get different results. They’re tapping into different aspects.

What do these results mean? Are both measures assessing the same attachment? Are these different aspects of attachment? The AAI assesses attachment-related states of mind through the organization and coherence of discourse; The ECR questionnaire assesses self-reported attachment anxiety and avoidance.

The ECR, ECR-R, and ECR-RS are all meant to tap into the trait – the style, the over-time pattern or predisposition. We developed the SAAM⁴ in the early 2000s when we needed a state measure and there wasn’t one. In our studies, we often use priming, but we never had a manipulation check to see if the priming actually worked. In a similar vein, in therapy – people might go through therapeutic changes, but there was no easy way to assess whether their attachment was changing.

So there was a need for a tool and the SAAM – the State Adult Attachment Measure⁴ – was that measure. It caught on. It’s been translated into a bunch of different languages, for different age groups. One of my colleagues in Belgium, Guy Bosmans, took the SAAM and applied it to youngsters, and developed state measures of other things too – state measures of trust, and of intimacy.

The big difference is: what are you focusing on? Are you focusing on, in the moment – change, progress? Or are you focusing on long-term trends and trajectories? And within overall, are you focusing on all relationships, or on a specific one? The ECR-RS separates it out – you have a general attachment style, then attachment to mom, dad, romantic partner, and best friend. And you can even ask: where does AI sit within that? We actually have an attachment-to-AI measure that we published in a 2021 paper.

Many people ask, Can I use your SAAM for my thesis or PhD? And I say: sure – but what are you using it for? Are you sure you’re looking for a state measure? Are you looking for changes, or are you looking for a trait? This is true for any trait, really. Are you thinking about whether you’re an introvert or extrovert in general, or are you thinking about how you’re behaving right now? That’s why measurement matters – what is your goal? Trajectories, and set points or fluctuations and changes?

Professor Gillath: Exactly. Exactly.

Professor Gillath: Talking about measurement of attachment, we notice another issue where people might use the term attachment but mean something else. Especially in the research of relational AI where people are coming from different disciplines – some from computer science, some from information systems, some from psychology or communication studies. Psychologists are relatively late to the party. And often, the people coming to this area from other fields, are lacking formal training in attachment and relationship science.

What happens then is that they’re using the same terms, but in very different ways and then measure attachment not using traditional tools or even the newer tools developed in psychology on the basis of attachment theory but their own developed tools. This can lead to what is called jingle-jangle fallacy – where people are using the same word, like “attachment,” but mean something completely different. I might be talking about relationships and social ties, whereas you might be talking about the file you attached to your email. Even when all the researchers refer to a social connection between a human and AI, it is important to differentiate between actually being attached to AI and the AI fulfilling attachment needs and functions (secure base and safe haven) to having a social connection to a chatbot without being attached to it, and potentially referring to liking, or trust rather than attachment.

People will say, Oh, these people are attached to their AI. But how do you know they’re actually attached to the AI? How did you measure it? What kind of measure are you using? Some people are basically ignoring the entire long history of attachment theory and research, and coming up with their own measures and their own conceptualizations. Which is a problem – because then we don’t know are these the same constructs? And if not, what kind of confusion can come out of that?

Most researchers who use the term attachment will use it to describe some social tie, but not every social tie is an attachment bond. People are throwing the word attachment around, but simply saying that someone is attached to an AI, doesn’t mean that the AI is actually fulfilling the needs or functions. This is similar to what Robert Weiss argued was the difference between social and emotional loneliness — only attachment can actually help relieve emotional loneliness.

Professor Gillath: I hope that we’re not too late. That’s my biggest hope. AI is introducing many challenges. Recently I saw an article about a new virus that was developed and created by AI. People are playing God here. Oh yeah, let’s release this new virus, see what happens.

We need to figure out how we regulate AI and we need to do it fast. And one of the things we need to regulate is how people are using AI in their relationships. Right now, there’s a push to regulate social media – from what age you can start using it, many researchers, educators, and parents say kids need to wait until they are 16 or older. I think the same should be true for AI, but even then, AI is built in such a way that even adults wouldn’t know what happened until it was too late. What I’m hoping for is that enough people figure out: look, this is dangerous. This can derail us. This can impact us as a species, not just as individuals.

We need to do the research, listen to the researchers, put the money there, and figure things out. What can we do to use it for better, rather than just tagging along with whatever the big tech companies are going to do?

If we have this conversation 5 or 10 years from now, I hope: A. we’ll still be here. B. we’ll still have close human-human relationships.

It is important to have hope and be optimistic, but we also need to be vigilant. We can’t just close our eyes and say, it’s just another technology, it’s like the phone or the computer. No – this is probably one of the biggest revolutions we’ve had. There’s a great podcast called The Last Invention, because their take is that AI might just be the last invention of the human race. We need to step into that with our eyes open wide, aware of the benefits but also the dangers – and fight for our relationships.

Professor Gillath: Disconnect.

Disconnect from technology and reconnect with humans. Go out into the real world – with all the friction, and the sweatiness, and the muddiness, and the scariness. Reach out. Connect with people. Say hi to the barista. Pick up the phone and call your mom. Reconnect with old friends. Join new groups. Have this human connection, because we need it.

A hug is something that is amazing. Connection with nature – being out in the woods, hiking with a group of friends, cycling with a bunch of like-minded folks. All of that is super healthy for us.

If more people do that – if more people connect, more people believe in love – I think the world’s going to be a much better place.