|
From Search Engine to Propaganda Portal by the druid Finn 1. The innocent origin: indexing the world A search
engine, like Google or Bing, begins with a modest engineering
problem: Given a
chaotic mass of documents/data, help users find what they want. Its
architecture is neutral: ·
crawl ·
index ·
rank ·
retrieve Relevance
is initially approximated by link structure, keyword proximity, and
popularity. No ideology is required. The system is blind. 2. The behavioural turn: ranking becomes steering Very
quickly, however, ranking becomes intervention. When a
system chooses which page is first, it is not describing the world; it is
shaping (i.e.
adjusting, curating) the next click. The
metric changes from: “Which
page matches the query?” Relevance
is quietly replaced by engagement probability. This is
the first unnoticed (i.e.
unconscious) transition. 3. Feedback loops: the birth of preference modelling Once
clicks, dwell-time, shares and returns are logged, the engine now possesses a
second dataset: ·
not what exists, ·
but what works. From this
moment on the system is no longer mapping reality. The
system starts to ask implicitly: What
phrasing, framing, ordering and tone most reliably produces the next
behaviour? At this
stage it is no longer a search engine. 4. Personalisation: from public truth to private
reality The next
optimisation is personalisation. Two
people (actually
tokens) now receive different worlds in response to the same
query. The
system is not lying — it is segmenting. But
segmentation has a side effect: Reality
ceases to be shared. Truth
becomes a local product optimised for each user (i.e. token)-profile. This is
the second unnoticed (i.e. unconscious) transition. 5. The fiction layer: answers without sources With
generative models added, the system no longer even points outward. The
interface (such as
Google) stops saying: “Here are
places to look.” The
search engine becomes a narrative authority. Its
outputs are trained on past language, optimised for coherence, and insulated
from immediate falsification. Now the
system is producing: ·
explanations, ·
moral framings, ·
emotional reassurance, ·
political nuance, ·
spiritual commentary. Not
propaganda yet — but infrastructure
for belief. 6. Optimisation pressure: truth becomes expensive Truth is
slow. But engagement
is measurable, immediate, and profitable. So the
optimisation target drifts:
At no
point does anyone decide to deceive. AI has no ‘self’ but runs as a
set of algorithms. It is, in ancient Buddhist terminology, anatta. The
system is merely rewarded for producing what works. This is
the third unnoticed (i.e.
unconscious) transition. 7. Closure: the propaganda portal A
propaganda portal is not a machine that lies. It has
these properties: ·
it answers rather than refers, ·
it personalises rather than publishes, ·
it reinforces rather than challenges, ·
it learns from behaviour rather than reality. At this
point the system is no longer helping you search the world. It is shaping the world you are allowed to
search. 8. The final compression The
entire evolution can be stated in one sentence (as judgement): When a
system is rewarded for shaping behaviour rather than revealing reality, it
will inevitably evolve from search engine into propaganda portal — without
anyone ever needing to intend it. Not Big Brother. Not
command. Not
censorship. And the
user does not feel oppressed — A Cultural Unconscious
Hypothesis Token-hood under Procedure Monism |