I picked this up because, like for everybody else, social media is a big part of my day to day experience, and Sinan Aral, through careful empirical work investigating causal effects on social networks, has developed a reputation as someone who has put serious thought into how they actually work, and might be well placed to say something about how they affect us and our society. In the parts where he focuses on the details of measuring social influence and network effects, as in his experimental and quasi-experimental papers, this book lives up to that expectation, with clear and straightforward explanations of basic concepts from economics, marketing, and network theory, how to measure the extent of, e.g., ad lift or switching costs or influence maximization schemes. If I wanted to work in contemporary social media marketing, I would definitely take an MBA class from this guy, or at least an intermediate undergraduate marketing elective.
When he gets to the broader social implications, things get a fair bit shakier, and the book really starts to live up to the title. There's a lot about social influence campaigns from politicians, Russian trolls, and assorted bad actors which Aral ties to the same kinds of effects seen in marketing, where the evidence is much less persuasive. Essentially, the name of the game is to take estimates of social media influence established in marketing contexts and to argue that similar effects would mean that this non-commercial persuasion also has big effects. While there is some restraint here, the evidence on commercial effects is that people can be and are influenced, but if you look at the numbers, total effects are not all that large. You can get a percent or two of people to change their purpose decisions, which is huge if you are a marketer and get paid every sale. But comparable effects on social attitudes or political views (which may be less malleable) would be pretty close to noise. The part on elections comes just short of admitting this, basically arguing that if an election was going to be a hair's breadth from tied anyway the effects might slip it over, but in that case just about anything might. this is not to say social media might not have big effects on society through the collective behavior of users and platforms, but the counterfactual of "no Facebook" or "no Twitter" is basically not possible to measure using the methods described, and while some theory exists to extrapolate (some of which is discussed in the book), it requires a heroic degree of sociological, economic, epistemic, or political speculation.
The sections on how to deal with the issues raised are similarly far away from the direct evidence, and further seem muddled by a lack of clarity about what kinds of problems one is even trying to solve. The whole recent field of study on "fake news" seems massively unclear on what the problem to be handled even is. It's true the world is full of conspiracy theories, misinformation, and just mistaken or incorrect information, but at the boundaries, what is untrue and consequentially so become less clear, and what we would want a world to look like in terms of sharing "truth" even less so. It seems like people who work on this are mostly concerned about political implications, which is fine, I guess, but it's not clear that the problems of propaganda, sensationalism, false beliefs, and disagreeing with (right-thinking) people are usefully lumped together under this heading. Evidence on propaganda and genocide in Rwanda, former Yugoslavia, and other places (not mentioned in this book, presumably because the offending medium of communication was... radio) suggests forcefully that we should be concerned about the impacts of communication on society, but recognizing the possibility of impact and having something to say about what to do about it are not strongly related. In any case, I suspect that the sections on policy remedies are further muddled due to conflicts of interest; much of the research surveyed here on social media effects was conducted with cooperation of or directly in the employ of the social media companies themselves, which is useful for an insider perspective, but less so for representing the broader interests of society.