[Corpora-List] Open Research Position (M.S. / Ph.D. / post-doc): Analyzing Routine Activities for Crime Prediction
Hardie, Andrew
a.hardie at lancaster.ac.uk
Wed Apr 23 21:19:44 UTC 2014
The best explanation I have seen on the drawbacks of any data mining for police work (textual or otherwise, pre or post deed) comes from writer/physician Ben Goldacre, who points out that the prior probability of any given purpose being a terrorist (or murderer or whatever) is so low that even the best data mining will produce results absolutely swamped in false positives. The same issue arises if you screen a whole population for some rare disease: the probability that a person with a positive result on the test actually has the disease is negligible.
http://www.badscience.net/2009/02/datamining-would-be-lovely-if-it-worked/
best
Andrew.
From: corpora-bounces at uib.no [mailto:corpora-bounces at uib.no] On Behalf Of Mike Scott
Sent: 23 April 2014 16:20
To: corpora at uib.no
Subject: Re: [Corpora-List] Open Research Position (M.S. / Ph.D. / post-doc): Analyzing Routine Activities for Crime Prediction
Seems to me much will depend on the beliefs and attitudes of the authorities as opposed to the experts. Looking for suspicious patterns is traditional in policing and this is just the same -- except powered by a turbo-charged mechanism. If the cops start to assume that the associations the software predicts are pretty well infallible, there will be no escape for the innocent and therefore this will be Minority Report (1984, The First Circle, etc.)
Fascinating thread (nearly missed because of a most misleading subject line)
Cheers -- Mike
On 23/04/2014 09:44, M.E.Sciubba wrote:
Beyond the legal aspect about the prediction of human actions and hence loss of basic civil rights, I think the key point here is the "cultural" variation (if any) of the values attached to what somebody tweets/sends through social media. It is a sociological analysis of the interaction between people(s) lives and (possible) threatening activities that is lacking, if I got it right...
Very interesting subject indeed, if I could only apply as a telecommuter post-doc ;)
Cheers,
Eleonora
____________________________________________________________________
Dr. Maria Eleonora Sciubba
Research Associate - Interactional Linguisuistics
Archivio di LInguA Spontanea
tel. +32 16 3 24795
cell +32 468177656
KU Leuven - Faculty of Arts
Department of French, Italian and Comparative Linguistics
Blijde-Inkomststraat 21, PO BOX 3308
B - 3000 Leuven
http://www.kuleuven.be/wieiswie/nl/person/00088846
________________________________
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm<https://owa.groupware.kuleuven.be/owa/redir.aspx?C=TdLzfiA_2UiPWm_Hku2lb346uUpNcdAINaOnchkCiVs0F2pBz-iwkuvyJDCbnn7U93SVwLpxYn0.&URL=http%3a%2f%2fwww.kuleuven.be%2fcwis%2femail_disclaimer.htm>
Be green. Keep it on the screen
2014-04-23 5:57 GMT+02:00 Alexander Yeh <asy at mitre.org<mailto:asy at mitre.org>>:
Zoltan Boka wrote:
Predictions are only as good as the initial data they're based on. In
this case that data could be incomplete or limited or biased (for
instance, lets say that one data point is the number of arrests made on
street x: lets also say that its police policy to harass and arrest
people living on street x- you can see how this can infect the process.)
But even if it can be done without bias and with a high accuracy, the
question remains- are some people literally destined to commit crime and
if so can that destiny be altered short of preemptively arresting them?
If you have a guess at a who is likely, you could concentrate undercover police to where (and when) that person tends to go and also to follow that person around. With enough evidence, you could also get search warrants and warrants to intercept their phone, etc.
Sent from my iPhone
On Apr 22, 2014, at 14:39, Adam Kilgarriff <adam at lexmasterclass.com<mailto:adam at lexmasterclass.com>
<mailto:adam at lexmasterclass.com<mailto:adam at lexmasterclass.com>>> wrote:
If a clever system can predict who is going to predict a crime - with
good, but far from 100%, accuracy, is the use it
a) rational policing practice
b) discriminatory
to use that information?
Seems to me, it's both.
Marek says
> But there are definitely many ways to abuse this technology as well.
I don't feel abuse is the main issue. Any use of it is
discriminatory. Should we trade off? Tough question.
Adam
On 22 April 2014 11:34, Marek Rei <marek.rei at gmail.com<mailto:marek.rei at gmail.com>
<mailto:marek.rei at gmail.com<mailto:marek.rei at gmail.com>>> wrote:
Here's an interesting article about how Chicago police is already
applying such technology (in somewhat troubling ways):
http://www.theverge.com/2014/2/19/5419854/the-minority-report-this-computer-predicts-crime-but-is-it-racist
I wouldn't say crime prediction technology by itself is evil, it's
more a question of how it's used. For example, I wouldn't have a
problem with a system that can prioritise a large list of likely
suspects after a crime has been committed, or is able to flag a
social media message calling for a hate crime. But there are
definitely many ways to abuse this technology as well.
Marek
On Tue, Apr 22, 2014 at 10:55 AM, Christian Pietsch
<chr.pietsch at googlemail.com<mailto:chr.pietsch at googlemail.com> <mailto:chr.pietsch at googlemail.com<mailto:chr.pietsch at googlemail.com>>>
wrote:
Hi Matthew,
so you want to build a heuristic precrime detector based on
routine
activities observed on social networks. Does that mean that
if, say, I
tend to update my status at the same time as some terrorist in
your
training set, your software will label me as a likely
terrorist and
put me on a no-fly list? Will I get a chance to prove my
innocence?
When you have some spare time, try to watch Minority Report.
Or did
this movie inspire your project? Then you must have
misunderstood its
message.
Your suspect
Christian
On Mon, Apr 21, 2014 at 11:34:11AM -0400, Matthew Gerber wrote:
> Hello,
>
> A new research position has opened within our lab, and we
are seeking M.S.,
> Ph.D., and post-doc researchers.
>
> One-sentence summary: We are mining social media for
indicators of
> individual routine activities for the purpose of improved
crime prediction.
>
> Longer summary: This project focuses on the spatiotemporal
prediction of
> localized attacks carried out against individuals in urban
areas. We view
> an attack as the outcome of a point process governed by the
interaction of
> attackers, targets, and the physical environment. Our
ultimate goal is to
> predict future outcomes of this process in order to increase
the security
> of human populations and U.S. assets and interests. However,
achieving this
> goal requires a deeper understanding of how attack outcomes
correlate with
> the routine activities of individuals in an area. The
proposed research
> will generate this understanding and in doing so will answer
questions such
> as the following: What are the dimensions along which
individuals'
> activities should be quantified for the purpose of attack
prediction? How
> can measurements along these dimensions be taken
automatically and with
> minimal expense (e.g., via social media)? What are the
implications of such
> measurements for attack prediction performance? Subsuming
these questions
> is the issue of geographic variation: do our answers change
when moving
> from a major U.S. city to a major U.K. city? There has been
plenty of
> previous work on spatiotemporal attack prediction (see our
Asymmetric
>
Threat<http://ptl.sys.virginia.edu/ptl/projects/asymmetric-threat-prediction>project);
> however, these basic questions remain unanswered, leaving a
> substantial gap in our understanding of attack processes and
their
> relationships with individuals' routine activities.
>
> More information can be found
>
here<http://ptl.sys.virginia.edu/ptl/projects/routine-activities-analysis-for-crime-prediction>
> .
>
> Sincerely,
>
> Matthew S. Gerber, Ph.D.
> Research Assistant Professor
> Department of Systems and Information Engineering
> University of Virginia
> _______________________________________________
> UNSUBSCRIBE from this page:
http://mailman.uib.no/options/corpora
> Corpora mailing list
> Corpora at uib.no<mailto:Corpora at uib.no> <mailto:Corpora at uib.no<mailto:Corpora at uib.no>>
> http://mailman.uib.no/listinfo/corpora
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no<mailto:Corpora at uib.no> <mailto:Corpora at uib.no<mailto:Corpora at uib.no>>
http://mailman.uib.no/listinfo/corpora
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no<mailto:Corpora at uib.no> <mailto:Corpora at uib.no<mailto:Corpora at uib.no>>
http://mailman.uib.no/listinfo/corpora
--
========================================
Adam Kilgarriff <http://www.kilgarriff.co.uk/> adam at lexmasterclass.com<mailto:adam at lexmasterclass.com>
<mailto:adam at lexmasterclass.com<mailto:adam at lexmasterclass.com>>
Director Lexical Computing Ltd <http://www.sketchengine.co.uk/>
Visiting Research Fellow University of Leeds <http://leeds.ac.uk>
/Corpora for all/ with the Sketch Engine <http://www.sketchengine.co.uk>
/DANTE: a lexical database for English <http://www.webdante.com>/
========================================
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no<mailto:Corpora at uib.no> <mailto:Corpora at uib.no<mailto:Corpora at uib.no>>
http://mailman.uib.no/listinfo/corpora
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no<mailto:Corpora at uib.no>
http://mailman.uib.no/listinfo/corpora
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no<mailto:Corpora at uib.no>
http://mailman.uib.no/listinfo/corpora
--
Be green. Keep it on the screen
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no<mailto:Corpora at uib.no>
http://mailman.uib.no/listinfo/corpora
--
Mike Scott
***
If you publish research which uses WordSmith, do let me know so I can include it at
http://www.lexically.net/wordsmith/corpus_linguistics_links/papers_using_wordsmith.htm
***
University of Aston and Lexical Analysis Software Ltd.
mike.scott at aston.ac.uk<mailto:mike.scott at aston.ac.uk>
www.lexically.net<http://www.lexically.net>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/corpora/attachments/20140423/09da50c1/attachment.htm>
-------------- next part --------------
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora
More information about the Corpora
mailing list