Insights from a new experiment suggest humans may still have the edge.
23 March 2026
By Emma Barratt
As it stands, the world leaves many of us lacking in real human connection. Some work long hours and don't have time to socalise, others are unable to do so due to physical limitations or just difficulty finding people like them. The lack of affordable third places also often thwarts the kind of random-chance connections enjoyed by previous generations. Short of an overhaul, loneliness seems likely to continue be a major problem for the UK population.
With chatbots "becoming increasingly embedded in social life," as Ruo-Ning Li of the University of British Columbia and team put it in their latest paper, there's potential for them to offer some kind of companionship. Whether or not that idea appeals to you personally, it feels somewhat undeniable that the accessibility and cost of an AI loneliness intervention would be a useful tool — that is, if it works.
To investigate the extent to which a chatbot can alleviate loneliness, Li and team put their own psychologically-informed AI chatbot, 'Sam', head to head with a random human to see which entity made participants feel less lonely in their first few weeks of university. Over the course of two weeks, 296 undergraduates participated in one of three randomly assigned conditions: chat daily with 'Sam', chat daily with a random unknown peer or, in the case of the control group, journal about their feelings.
By comparing surveys on loneliness conducted both before and after that 14-day window, the researchers found that only those who had spoken with an actual human felt significantly less lonely once the intervention was complete. Those that chatted with 'Sam' felt no less lonely than those in the control condition. A look at daily reports provided by participants also revealed that only those who chatted with a human felt less isolated and reported more positive emotions than controls once those two weeks were up.
The chatbot condition did, however, see a few wins; compared with the control group, both the AI and human chat partners helped to reduce negative emotions, and people felt equally close to both the chatbot and human partners. When participants were told they could keep chatting beyond the two weeks, though, only 13% of those having an exchange with 'Sam' chose to continue. When you compare this with the 33% that chose to continue chatting with their new human conversation partner, it seems that at least in this situation, humans continue to be more appealing to talk to.
The team suggests that this difference might in part be due to limitations in expressions of empathy from 'Sam'. "While we designed our chatbot to express high levels of empathy, it is possible that in practice, it often failed to do so." The team asked GPT to rate the empathetic qualities of the conversations participants had, but due to this AI model being a 'black box', it's unclear whether these assessments included in the paper are valid, or subject to hallucinations.
Even so, this study suggests that humans still have the edge in providing a sense of connection, even when they're just a random stranger. Previous research has suggested that perhaps the artificial nature of the AI chatbot could have left participants feeling 'empty' while chatting with them — something that isn't easily remedied by tweaking algorithms. Whether further shifts in cultural attitudes towards chatbots, evolution of the technology, or similar factors will change that in future remains to be seen.
Read the paper in full:
Li, R.-N., Folk, D., Singh, A., Ungar, L., & Dunn, E. (2026). Is a random human peer better than a highly supportive chatbot in reducing loneliness over time? Journal of Experimental Social Psychology, 125, 104911. https://doi.org/10.1016/j.jesp.2026.104911
SOURCE:
https://www.bps.org.uk/research-digest/chatbot-vs-random-human-which-better-tackling-loneliness(accessed 14.4.26)
No comments:
Post a Comment