The past few years have seen an increasing polarization of western politics and a string of unanticipated election results. This division has been explained by some as the product of informational ‘echo chambers’, where people are confronted only with opinions and information that reaffirm their existing beliefs and are ignorant of the public mood elsewhere. Echo chambers may exist in online social media, within a population of newspaper subscribers or within people's local social or religious communities. Echo chambers have been implicated in recent unexpected political outcomes such as the vote for Brexit (http://go.nature.com/2ndbFhW), the rise in popularity of Donald Trump (http://go.nature.com/2nBDLUN) and of populist politicians in several European countries (http://go.nature.com/2mPW4Bj), and the apparent blindness of the UK Labour party to its crushing general election defeat in 2015 (http://go.nature.com/2nRVrJi). For scientists, the particular concern is that seemingly intractable positions in public debates, such as what action to take on climate change or the ethical regulation of biotechnologies, may be maintained within echo chambers.

The potential consequences of echo chambers are worrisome at a time when cross-political understanding seems to be at a low ebb.

While echo chambers as a concept have gained considerable traction in the media, in a Comment in this issue of Nature Human Behaviour, Margetts (article number 0086) reminds us that the scientific evidence for an outsized influence of echo chambers in modern society remains equivocal. Even in this age of big data, large-scale studies into echo chambers are rare.

Credit: TAMAS NEMETH / ALAMY STOCK PHOTO

Also in this issue, Shi et al. (article number 0079) leverage an impressive dataset to present empirical evidence for a potential echo chamber effect within public consumption of science literature in the United States. In a network analysis of millions of online book purchases, the authors show that conservative or liberal political books are purchased alongside science books in distinct subject areas. While ‘red’ (conservative) books tend to be bought alongside books on applied sciences, such as geophysics and criminology, ‘blue’ (liberal) content is bought alongside books on basic science, such as physics and astronomy. Blue books were also co-purchased across a wider range of sub-disciplines, while the red selection was for a more narrow range of science. These findings point to a divide in interest and exposure to scientific information, mediated by political alignment, and provide an example of how researchers may go about searching for political informational echo chambers.

The potential consequences of echo chambers are worrisome at a time when cross-political understanding seems to be at a low ebb. More research of the kind conducted by Shi et al. will be needed. However, as Margetts discusses, research efforts may be hampered by lack of access to data and, in the case of online social media, lack of transparency around the marketing algorithms that direct content to users. Given the global penetration of social media into people's lives and its potential as a platform for further segmentation of public discourse, identifying means or incentives for greater transparency and access is a priority. Furthermore, future research will need to combine the broad-scale picture provided by big data studies like Shi et al. with methods that can tap into individual-level decision-making. In this way, researchers can identify and map suspected echo chambers and elucidate what influence they truly have on individual and collective behaviour.