This book can be summed up in two sentences: "People instinctivly react to artificial interactions (with computers and media) the same way they react to interactions in real life. This can be used to manipulate people."

Except for the authors' pro-manipulation bias, this media=reality "equation" is nothing new - Steven Pinker talked about "new media and old brains" in his first book, published in the 1980's. Ted Nelson said that users instinctivly interact with computers "naturally" in "Thinkertoys", published in the 1970's. Even Arthur C. Clarke mentioned the phenomenon in "3010."

In this book, the authors report on previously done social science research they have replicated and repackaged with a popular, marketing spin. Academics working in the social or cognitive sciences will find nothing new here. The authors enjoyed a brief heyday among dot-commers, but these were the same people who were buying $500 office chairs.

Don't bother.

The media equation, as introduced by Nass and Reeves, is that "media equals real life" and that our interactions with media are "fundamentally social and natural" (p. 5). This book is a popularization of established, replicated research on how people interact with television advertising, tutoring systems, error messages, loud noises, sudden movement, etc. For instance, one widely replicated result is that computer tutoring systems get better evaluations if the evaluation program is run on the same computer. Moving the reviewer to a new computer (with the same program), significantly lowers the score. The social science literature shows that teachers who collect their own evaluations score much more highly than those whose evaluations are collected by others. This is the kind of evidence Nass and Reeves bring to bear in support of the media equation. They don't claim that we are consciously thinking about the computer's feelings and don't want to hurt them. Rather, to the contrary, subjects claim they were doing no such thing. Yet the evidence of our behavior seems incontrovertible.

The media equation is a good enough predictor of user behavior, at least for telephone-based spoken dialog systems of the form my company builds, that it has informed our designs from top to bottom. Our applications apologize if they make a mistake. Callers respond well to this. Sure, the callers know they're talking to a machine, but this doesn't stop them from saying "thank you" when it's done or "please" before a query or feeling bad (or angry) if the computer can't understand them. Another strategy recommended by Nass and Reeves that we follow is trying to draw the caller in to work as a team with the computer; again, Nass and Reeves support this with several clever experiments. There is also a useful section on flattery, looking at the result of the computer flattering itself and its users; it turns out that we rate computers that flatter themselves more highly than ones that are neutral.

Among other interesting explanations you get in this book are why we're more tolerant of bad pictures than bad sound, why we focus on moving objects, speaking rate equilibrium, what we can do to make someone remember an event in a video, and the role of gender.

This book is very quick and easy to read. I read it in two days while on vacation it was so fascinating. In contrast to the classical yet dry social science format of hypothesis, experimental methodology, results, and essentially a summary of the results as a conclusion, Nass and Reeves only vaguely summarize their experimental methodology and take a no-holds-barred approach to drawing conclusions. This may annoy social scientists, most of whom expect their own kind to be far more circumspect.

This book is an absolute must-read for anyone designing mediated interfaces. For those who don't believe the results, I'd suggest running some experiments; our company did, and it made us believers.

Reeves and Nass find that people use social conventions towards intelligent systems. Without further thinking they decide it is something new, something great. We are thinking of interactive systems as other social beings is their explanation. With some imagination you could come up with a number of other explanations, that equally well fit the data.

I do not disagree with their findings, but I really disagree with their conclusions, especially the eagerness and determination with which they jump to them. However I notice their ideas seem pretty convincing, and here lies my real worry with this book. So if you decide to read it anyway keep asking yourself if the conclusions Reeves and Nass jump to are really as worthwhile as they make them appear.

This book is one of there rare ones that really is based on scientific material. Reeves and Nass make their extremely interesting and useful observations on how users react to media available to the HCI professionals in an intelligent and intertaining fashion. Their findings are of great value to anyone engaged in computers and media. The book is not a how-to-do guide, but at the end of each chapter you will find useful recommendations on how to help your users/viewers feel more comfortable with your product.

I highly recommend this book.

A previous review called this book "psuedo-scientific drivel."

In fact, this book is far from it. Well, as far from it as social science can get. In fact, is the most "scientific" of the user interface books I have read.

The main point I took away from the book is that people interact with objects, especially interactive and media devices, as if they were people. They demonstrate that when user interfaces are designed to be polite and interact in a positive social manner, the person has a much more enjoyable and profitable interaction.

Other books on the topic of user interface design are far less scientific, relying on generalizations and suppositions rather than constructing a study. Some use data from a usability evaluation, but these are often far from scientific.

The authors construct hypotheses, usually based on the results of studies of interaction between humans, and see if the results of the results hold true for human-machine interaction.

Usually, it does.