[LINK] New MIT Big Data Course

Bernard Robertson-Dunn brd at iimetro.com.au
Fri Jan 10 17:02:32 AEDT 2014


On 10/01/2014 4:37 PM, Roger Clarke wrote:

> That's good, but it addresses just one small threat among many, viz.
> access by a third party that hasn't paid to join the club.
>
> There's nil sense of such aspects as:
> -   the data's original purposes
> -   semantic inconsistencies
> -   data quality factors (datedness, completeness, etc.)
> -   expropriation and consent
> -   copyright
> -   deidentification
> -   reidentifiability
> -   constructive data falsification in order to deny administrative value
>       while sustaining statistical value
> -   checking of inferences drawn against the real world
> -   transborder issues
> -   extra-territorial reach of law (which the US asserts in spades,
>       but the EU in particular is slowly learning to claim as well).
>
> If those kinds of issues aren't addressed, then CSAIL is just
> teaching people to be (and are just acting like) glorified plumbers.
> Even in their own terms, the problem statement is wrong, and hence so
> is the solution spec, and so is the answer that tumbles out of the
> exciting algorithm.

All these issues are highly relevant regarding the management of the data.

I'd like to add another potential flaw to the big data movement - yet 
another of those wildly optimistic, technology driven crazes.

Data cannot exist without models. Models that explain existing 
relationships and/or predict future data. Proponents of big data suggest 
that models will "emerge" from analysis of the data. This is bunkum. The 
biggest single problem that big data analysts will have to face is 
sorting out cause and effect - traditionally not always an easy task. 
There needs to be a high degree of expertise in model creation, 
manipulation and understanding - things glaringly missing (at least to 
me) from the MIT course.

As Roger points out - the problem statement is wrong - in fact I'm not 
sure if there is a problem statement, just a claim that "big data" is a 
fantastic solution.

And when I see phrases like "share, engage, and ideate with other 
participants" my immediate reaction is Ideate?? Ideate is not a new 
word, but it is a bit archaic - or maybe its just trendy to use it.

BTW, it means to form an idea, conception, or image of.

-- 

Regards
brd

Bernard Robertson-Dunn
Sydney Australia
email: brd at iimetro.com.au
web:   www.drbrd.com
web:   www.problemsfirst.com
Blog:  www.problemsfirst.com/blog




More information about the Link mailing list