|
John,
Good of you to comment. I agree with what
you say about being up front about collecting data on email traffic,
particularly now that I understand the web long technologies and
the issues better. At first, I simply assumed that there would be no
privacy problem as long as the survey was anonymous, no addresses or personal
data were recorded, and all statistics were descriptive and at the summary
level. I thought of it more as a non-invasive measurement like measuring
the wear of floor tiles in a museum to get an indication of traffic. But,
that still leaves the question of measuring something without explicit
consent.
This is an interesting issue. I am going to
consult the human subject protection people at the University of
California, Berkeley for a history and an assessment of these issues.
I am also curious regarding commerical applications. If counting
events and tracking activity on websites without the explicit consent of
the people involved is unethical, then what are web log analyzer companies
and software like Analog and Surfstat doing? Does the distinction
between ethical and unethical boil down to embedding things like transparent
images to enable measurements without informing recipients? I wonder if
visitors in public places are aware that an infrared beam is counting
them? In research, part of ethical issues involves
benefits. The more I think about this, the more I am of the opinion that a
scientific study that requires transparent tracking of email should go through a
human subjects protocol process.
In the meantime, if we proceed we are thinking
about two possibilities: The first is to ask recipients to let us know if
they pass along the invitation to participate and to how many. The second
is to explain to recipients what we are doing (counting) and for what purpose
and ask for their consent. Both pose significant problems of response
rates and the reliability and validy of results, but that's the name of the
game.
As I said to Stil, I am delighted to see people in
the Internet analysis business who are sensitive to privacy issues.
Thanks for your input.
Larry Boyd
----- Original Message -----
Sent: Monday, May 17, 2004 6:18 AM
Subject: Re: [analog-help] Email Opening
Count and Forwards for Research on Pyramiding in Public Advocacy
Campaigns
Lawrence, Why hide the
fact that you are collecting data? Since you are trying to be up-front
about things. Be honest with the users and you might get what you want.
John
|
| Stilgherrian <[EMAIL PROTECTED]>
Sent by: [EMAIL PROTECTED]
05/14/2004 08:16 PM
Please respond to analog-help
|
To:
"LAWRENCE BOYD" <[EMAIL PROTECTED]>,
[EMAIL PROTECTED]
cc:
Subject: Re: [analog-help] Email
Opening Count and Forwards for Research on Pyramiding in Public Advocacy
Campaigns |
At 15:17 -0700 14/5/04, LAWRENCE BOYD wrote:
>I am a social
researcher doing a study of email pyramiding
>strategies for
citizen-initiated polls. We have a url for a poll in
>an email
invitation to participate. [snip]
>Can we get this [following]
information using Analog? (Somewhere I
>read that you can embed a
transparent image in the email that would
>produce a request that could
be counted showing that the email was
>opened?)
Lawrence, the
basic issue is that Analog can count and measure what
is in web server log
files -- nothing more, nothing less. So whatever
you do must somehow
generate a request for a file from a website.
That request would then be
logged, and Analog can slice and dice that
data in all sorts of
ways.
Any web log analysis program can do this, the ability is not
unique
to Analog. It's all about how you set up your email, not about how
you analyze the logs.
Yes, you can embed an image in an HTML (that
is, a formatted) email,
and have that email come from a web server rather
than being attached
as part of the email. If that image is loaded as the
email is read,
then a request to the web server is generated and logged.
Indeed, if
you generate a unique ID number for each email you send out --
the
link to the image could be of the form
http://your.server.com/someimage.gif?55449582754, with a different
number in every email sent -- that number is also logged. You can
then
see which email addresses generated a request, and you will know
that
duplicate numbers can be traced back to the initial address.
There are
also tricky things you might do with _javascript_ or other
scripting
languages to generate a *new* unique ID when the email is
opened, thus
enabling you to trace the email as it's passed on to
others.
However, there are two major problems with this
technique:
* It will not deliver accurate results.
* Its ethics are extremely questionable.
It won't deliver accurate
results because not all email client
programs will load the
image.
* Many people (though probably in a minority these days)
use email
clients which do not render HTML email. They will
only see messy
HTML code, and are unlikely to follow the
image link manually to
load the image. These people won't
appear in your count.
* Many people whose email client *does*
display HTML email will
have the display of external images
turned off -- because it's
considered to be an invasion of
privacy for someone to track
whether they open an email or
not, and when. Indeed, this is
exactly the technique used by
spammers to validate whether an
email address works or
not.
The number of people blocking such external images
already high
and is increasing as email programs improve
their security and
privacy.
* Similarly, any
scripted tricks are increasingly likely to be
blocked,
because running unknown program code as you open an
email is
a security risk. This is exactly how virus propagate.
The method is
ethically questionable because:
* The external image-load
technique does not allow the user to
provide informed consent
to having their behaviour logged before
that logging takes
place.
* Potentially, if you do the track-the-hand-on stuff,
you're
compiling a list of who's a friend of whom (at least
via their
email address, and that's fairly easy to match back
to people)
and, given the context of your research question,
matching that
to their political beliefs. How is that data
going to be handled
and people's privacy
protected?
So, to answer your specific questions...
>We need to learn:
>1. How many
people opened the email,
No, because many people will not generate a
request to the web server
even if they do open the email.
>2.
How many clicked on the url, and
If you mean how many manually
clicked on the URL to your poll, you
already know this from your web
server logs. Or at least you know how
many requested the web page with the
poll on it (because you're
logging it), and then how many completed the
poll (because you're
also logging that).
However, there's no
completely reliable way to know how they came to
find that web address
unless you add that unique serial number to
every email. But you won't be
able to analyze the pyramid down past
the first level.
>3.
How many forwarded the email to their friends - and the number
>of lines in a chain, if possible.
No, for the reasons outlined
above. At least not reliably. And at
least not ethically.
Of
course, if you threw ethical issues out of the window, this sort
of thing
is possible. And indeed it's done all the time by people
like
Emode/Tickle, and the less-than-reputable folks who embed
"spyware"
programs into people's email clients under the guise of
providing them
with "cute smiley faces for their email". Such folks
will quite happily
track every email a person sends, who they send it
to and when, compile
comprehensive behavioural profiles of
individuals and sell
them.
But since you're researching "public advocacy campaigns", and
this
sort of hidden user-tracking represents an attitude which is the
exact opposite of advocating the rights of the public, I'm hoping
this
mis-match prevents you choosing this path for your research. :)
I hope
that helps,
Stil
--
Stilgherrian
<[EMAIL PROTECTED]>
Internet, IT and Media Consulting, Sydney,
Australia. ABN 25 231 641 421
mobile 0407 623 600 (international +61 407
623 600)
fax 02 9516 5630 (international +61 2 9516
5630)
+------------------------------------------------------------------------
|
TO UNSUBSCRIBE from this list:
|
http://lists.isite.net/listgate/analog-help/unsubscribe.html
|
|
Digest version: http://lists.isite.net/listgate/analog-help-digest/
|
Usenet version: news://news.gmane.org/gmane.comp.web.analog.general
|
List archives:
http://www.analog.cx/docs/mailing.html#listarchives
+------------------------------------------------------------------------
|
