From: Jason W Solinsky <solman@MIT.EDU>
Date: Sun, 21 Aug 94 22:35:39 PDT
To: cypherpunks@toad.com
Subject: Achieving Privacy while Enabling Marketing
Message-ID: <9408220535.AA09679@ua.MIT.EDU>
MIME-Version: 1.0
Content-Type: text/plain


[I sent a variant of this to several lists earlier this year. It was
originally a response to a post on online-news asking what the future of
advertising will be like. It suggests a system whereby marketers only
get the information that they absolutely need and pay for what they get.]

[The software is being debuged and will be ready "soon". I had thought I
would be alpha testing it already, but its going to take me atleast another
two weeks and quite possibly much more. I'm working on it :( ]

Well here is my view of advertising, supported by a product that I will be
indroducing latter this year:

In the NEAR future, content in the new media will be nearly entirely
divorced from advertising. Online media, be they magazines, interactive
games, fiction, discussion groups, or simple news, will be paid for at
full cost by the consumer. At the most, some of this content will contain
hints about where advertisements should go and keywords suggesting types
of advertisements which are particularly apt.

[herein, when I say browser, I mean the instrument by which the user
examines the content]

The browsers that the reader/viewer uses to read/view/listen to the medium
will pay the cost of the content from the user's electronic account almost
transparently. The browser will then schedule advertisements to pay for the
cost of the content.

The browser has precisely one valuable resource, the attention span of its
user. Periodically [the frequency is established by preferences that the user
can set] the browser will conduct an auction. All the agents that are
interested in taking up some of the user's time [either to deliver an
advertisement or ask a question] submit a "profile" to a third party that
is mutually trusted by the advertisers and the browser. The user also
submits a profile.

The agent's profiles contain two things:

A) a routines which, given the characteristics of the user, returns a price
   that the advertiser is willing to pay for the user's time.

B) the set of conditions which the advertiser is willing to agree to, and how
   much less it is willing to pay if it has to agree to those conditions. For
   example, suppose the user is a citizen of a cyberspatial government that
   prohibits fraud. If it is a popular government, the agent will have done
   one of two things:
   
       1) It will have been pre-approved by the cyberspatial government's
          censors [who analyzed it to make sure that it wouldn't confuse
          the netizens who support that government] for either a one-time
          fee or a per use fee or both.
       
       2) Its owner will have given the government a deposit from which to
          subtract fines (using a protocol that supports the use of a third
          party adjudicator).
       
   After doing this the agent will have received a certificate [probably one
   that requires a time sensitive number from the government to demonstrate
   its validity. There are many protocols for doing this. This allows the
   government to collect its "per use fees". Now acquiring this certification
   has cost the advertiser money, so the profile includes information on how
   much less the advertiser is willing to pay if the user requires agreement
   to the anti-fraud laws of the government in question.

The user's profile will contain:

A) the answers to the questions that have been asked of it. These are
   completelly uncertified.

B) certified characteristics. These may simply be answers which have been
   certified by third parties as time invariant [make sure that the user
   doesn't change his answers all of the time] or they may be certifications
   by vendors that the user spends X dollars on product Y each year, or they
   may be certifications by a third party that the user has bought a gift
   certificate for a class of goods. [For example, certification that a user
   has bought a $100 gift certificate for any truck in class X for Y dollars,
   something that certainly is worth something to truck advertisers.] All
   users will also have several certified characteristics describing how
   frequently they look at advertisements.

   Most of these certificates also require a fee for verification. So the
   user will not be able to demonstrate the certificates without first paying
   the certificate issuers a small fee. [Presumably, in order to fight the
   false issuance of certificates there will be anti-fraud agencies which
   in turn also ceritify the parties issuing the certificates.]

C) The user's preferences. Users may require more money for femenine hygene
   commercials than for beer advertisements. Users may also pay extra for
   advertisements that promise to be entertaining. This promise would
   be backed up by an agreement that requires less entertaining
   advertisements to pay a penalty while more entertaining advertisements
   get a discount. After these commercials the user would rate them. The
   algorithm used to determine who gets what would keep the amount that the
   user pays constant.

   Presumably the algorithm would be tweaked so that non-commercial
   entertainment could be sold through the same medium. Jokes could
   show up and promise to be funny [there could even be certification
   agencies to verify this [the comedian's guild]]. The user would
   then rate the jokes and sufficiently funny jokes could make money.
   [Presumably jokes certified by highly acclaimed comic certification
   agencies would only show themselves if the user were giving them a
   price at which they could make money].
   
   The governments to which the users belong would also be in this part.
   For each law that is important to the user, there would be an additional
   amount of money that the user is willing to pay for advertisements that
   comply with that law ranging from zero to infinity. Children's
   restrictions also belong here. The user could be subscribed to a law
   which gives a near infinite credit to those advertisers that can verify
   that they are suitable for children.
   
   Finally, users [and advertisers but I forgot to mention this earlier]
   could pay extra for advertisements that are in context. If the user
   was reading content about X and there is an advertisement related to X,
   the user or advertiser could be willing to pay extra to get an
   advertisement that is relevant to the current content.

The third party puts all this information together and for each advertiser
determines how much he is willing to pay. Then, it modifies the amounts
the advertisers are willing to pay by the credits that the user offers.
Finally these amounts are normalized by the resources that they require.
30 second commercials need twice as high an amount as 15 second commercials
and many times more than visual pop-up and streamer commercials. The highest
amount wins... but the highest amount pays what it was willing to (i.e. the
credits that the user offers are just for determining who wins, not how much
they pay.

Depending on the content they watch, the cost of their net services, and
their advertising preferences, some people will make money just by reading
or viewing the screen while others will have to pay.

The question of guaranteeing advertising time has come up. The way this will
be taken care of is by establishing attention futures markets. Agents will
be able to purchase futures on somebody's attention. So if coke wants its new
advertising campaign to role out to X% of the population at the same time,
it can buy futures. The futures will give coke the right to show its
commercial to a certain person at some point in time between two times.
If the commercial can't be shown, a penalty of Y cents will be paid to coke
by the issuer of the future. The futures would be traded my arbitrage agents
that think they can model the user's advertisement viewing patterns and by
the browser which has an advantage as an insider.

Cheers,

JWS