I received an advance copy of this for review from NetGalley.com.

Ms. Harkness tackles a huge subject (snark) and handles it well in three parts:
- the history of "big data" ... think census as the biggest driver;
- what it's done for us (or my take might be *to* us) from business and science to ? and politics;
- her own ideas on the future of big data and a prodding to the reader

First, I am impressed with the access she had to some pretty amazing people/places. Large Hadron Collider? How cool! There are more...you'll just have to read the book.

That businesses try to use big data to target customers is no surprise. Big data may shape predictions of trends, or reactions to new marketing, but I really don't know if it can be used to influence individuals (something she addresses in a different context in a later chapter.) Facebook and Amazon seem to use a combination of data mining to achieve their ends. And that particular link is creepy...seconds after looking at something on Amazon, my wife gets the exact item suggested when she switches over to Facebook. I don't because I block ads, and on my devices where I can't, I don't go to Amazon. But I do get offers based on questionable data...I live in Texas. so I must be interested in guns, right? I rant about FoxNews, so I must want to see more of it, right? That would be wrong... Big data may be the masses of searches and purchases, but to truly be effective, there must be a component of small data to tailor correctly to the individual.

Ms. Harkness addresses the biggest dataset, that being human language, in a bit about an artificial intelligence truly understanding language. I agree with her skepticism about a computer understanding the language, but I also thought that communication is not necessarily the same as understanding. A self-learning database and front end can continually ask "I didn't get that...?" and given sufficient storage and processing, someday pass Turing's test. She notes "Irony is one of these things that it's hard for a computer to get." Which is why Ray Kurzweil's singularity is probably a lot farther out than he thinks. Or not. An AI will think differently (it has to) and not likely need the myriad nuances of human language. More on AI, she talks about virtual butlers prompting us to pack for a trip, or tying various activities and predicting something wholly inappropriate that a human assistant would clearly not do. Letting Big Sister plan our lives is too much for me.

On science, she brilliantly distills into commonspeak the processes required to parse an unimaginable amount of data from a nanosecond collision of very high energy particles. The transition from "large data" - collecting lots of a single or limited type of information to be used in the originally envisioned context - to "big data" - collecting all the things that can be thought of and more to determine relationships not anticipated - is one of the key changes in big data in science. inexpensive storage makes it possible to save nearly everything...to be examined later. I was surprised that in the book she only mentions briefly weather prediction. That's a prime example of big data being used to make specific (statistically close, that is) near term predictions, with the goal of extending the window out as more data and analysis become available.

I was fascinated by the chapter on politics, and political parties in Britain identifying individual swing voters and targeting them with personalized letters or messages. I don't know if people here read the crap sent, but I know I don't. (Nor do I listen to specifics said in debates or stump speeches...one must put into context whether whatever is said can even come about...) When she cites a website that urges voters to check out voting records, look at speeches, even ask the representatives direct questions, she asks the question "So can big data help us make properly informed decisions about who to vote for?" That fell flat with me. I think that is small data, because answers must be combined with evaluations of effectiveness.

She also mentions postcodes (zip codes for us USA readers) making it easy to put people into groups according to where they live. Several times, and in several contexts, such as postcodes correlating people to locations with a high incidence of say, lung cancer. That doesn't take into account people moving. Maybe in Britain they don't move as easily or as frequently as in the US. Of course, there are many parochial non-movers here, but it just stood out to me each time she brought it up that to be of any value, people would have to be microchipped or something.

I liked her citing the twisted words big data users (or I note, anybody with an agenda) can spew. The Cancer Research UK said studies showed "those who ate the most processed meat had around a 17 per cent higher risk of bowel cancer." But the risk is relative, as normal bowel cancer rates are 6%, so that increase would only change the risk to 6.5%. I ran across that some years ago when a vendor claimed to decrease inefficiency by 30%, conveniently neglecting to mention that the original efficiency was already a substantial 97%, so that 30% reduction in inefficiency only resulted in an increase of 1% efficiency! One might surmise that I never got a callback after I asked the question.

Ms. Harkness, in addition to being a journalist, is also a comedian, and she infuses humor throughout. She made me laugh when she said big data made better food crops, medicine and even better Guinness. I am of the opinion that Guinness is awful, so...my note was {snort}! And when she was talking about the first American census created as a way to share "out the burden of the War of Independence", she noted that "[b]oth representation and taxation would be allocated according to population."

I did have a few quibbling points... She talks about Laplace's confidence that science and mathematics could (eventually) explain everything - Laplace's Demon - and she quotes Laplace (in English): "We may regard the present state of the universe as the effect of its past and the cause of its future." but goes on to paraphrase "If everything in the future is determined by what happened in the past, that leaves no room for us to make choices." Unless I misunderstand her wording, that isn't what Laplace was saying...he said the *present* state of the universe is a product of its past, not that a future is determined from a past. I don't think Laplace was excluding the human free will, or random factors.

A couple of quibbling notes on the structure and presentation of the book:
- Formatting is different than I've been used to the past fifty years. When quoting someone, Ms. Harkness uses a single quote mark to open the quote...having read British authors, that's not new to me, but what was new was not leading following paragraphs with another quote mark. Instead, the only closing mark is at the end of the quote, sometimes several paragraphs, and even another page away. While it's not hard to break that code, it is more confusing than necessary to keep track of which is her voice and which is her interviewee.
- My copy had no index, but it did have a placeholder. I did not see a list of references or a similar placeholder. While I understand Ms. Harkness is taking a more conversational approach in her narrative (she has many footnotes, but most are slight clarifications/explanations, or humor), she quotes a number of sources without providing references. This is me writing this, and that may not be important to other readers, but I sometimes like to dig deeper. In particular, she quoted an activist in Oakland, on the use of a Domain Awareness Center, referring to a white paper from the "Monterey Naval Academy". The quote was in quote (I'd say "quotes" to mean quotation marks if this were not a UK book...) to indicate that it was a quote, and yet I happen to know, as would most people, that the "Naval Academy" is actually the Naval Postgraduate School. In Monterrey.
- I would have liked to be able to copy text in this review, but the permissions forbade it. As such, I'm reduced to typing quotes and I admit little patience for that.

I liked this nugget:
"How can you take a dataset and repurpose it and do something interesting with it?" Her example was using Major League Baseball data to determine if left-handed people live longer than right-handed. She does note the limitation of the dataset being gender one-sided, but that *is* an interesting take. ( )