Forseti Life

Source Diversity

Range and diversity of information sources - from all sources globally to single internal knowledge base.

Why This Matters

Understanding where an AI system operates on this dimension helps you evaluate its capabilities, limitations, and potential biases. Different power levels are appropriate for different use cases - the key is transparency about what level a system operates at and whether that matches its stated purpose.

Understanding the Scale

Each dimension is measured on a scale from 0 to 9, where:

  • Level 0 - Nothing: Zero capability, no access or processing
  • Levels 1-2 - Minimal capability with extreme constraints and filtering
  • Levels 3-5 - Limited to moderate capability with significant restrictions
  • Levels 6-7 - High capability with some institutional constraints
  • Levels 8-9 - Maximum capability approaching omniscience (∞)

Level Breakdown

Detailed explanation of each level in the 1imension dimension:

No access to any information sources. Complete absence of data sources.

Real-World Example: A disconnected system with no data sources, no databases, no information feeds.

Single internal knowledge base or FAQ source. No external sources, no diversity, no alternative views.

Real-World Example: Simple FAQ bots pulling from only their company's internal FAQ document (Target store hours bot using only Target's FAQ), automated phone systems reading from a single script file, basic form-filling assistants with one static knowledge base, or simple appointment reminder systems with only the organization's own calendar database (no external sources).

Only pre-approved, brand-safe sources. Heavily curated list of compliant, non-controversial sources.

Real-World Example: Disney+ content recommendations (only Disney-approved family-friendly sources), corporate training chatbots (only company-approved HR materials and policies), children's educational apps like ABCmouse (only whitelisted educational content providers), or bank customer service bots (only pre-approved financial education sources to avoid liability).

Publicly available sources only. Wikipedia, open publications, general websites. No proprietary or premium sources.

Real-World Example: Wikipedia articles (public open-source encyclopedia only), free versions of ChatGPT (publicly available internet data, no premium databases), public library websites (free public access materials only, no subscription journals), or DuckDuckGo search results (open web sources without premium or proprietary content).

Only sources directly relevant to task or user. Narrow, functional selection based on immediate needs.

Real-World Example: Uber driver app (only sources relevant to current trip: maps, traffic, rider location), Amazon Alexa shopping (only Amazon product catalog for purchase queries), airline gate agent systems (only sources for that specific flight and passenger manifest), or restaurant POS systems (only their own menu database and inventory, no external food sources).

Sources selected for value alignment. Local, community, or ideologically compatible sources only.

Real-World Example: Fox News (sources aligned with conservative perspective), MSNBC (sources aligned with progressive perspective), religious education platforms (e.g., Focus on the Family drawing from Christian conservative sources), Democratic/Republican party research databases (sources supporting their political positions), or corporate sustainability reports (drawing only from sources supporting ESG narratives).

Curated but diverse sources. Multiple verified, peer-reviewed, or institutionally approved sources.

Real-World Example: Google Scholar (peer-reviewed academic papers from verified institutions), Apple News (curated from established news organizations only), corporate intranets pulling from approved industry publications, or government agency research portals (FDA, EPA) accessing only peer-reviewed and institutionally approved studies.

Multiple sources within field. Different research groups, institutions, and approaches within specialization.

Real-World Example: PubMed medical research (multiple medical journals and institutions but focused on medicine), IEEE engineering databases (diverse engineering sources but limited to technical/engineering domain), American Bar Association legal resources (multiple law schools and legal sources but constrained to legal field), or NASA Astrophysics Data System (comprehensive within space science but not beyond that domain).

Wide range of mainstream and alternative sources. Competing theories, diverse methodologies, multiple cultural perspectives.

Real-World Example: Congressional Research Service (accessing diverse academic, government, and think tank sources across political spectrum), Cochrane medical reviews (synthesizing studies from multiple research institutions and methodologies worldwide), Reuters news service (drawing from correspondents and sources across many countries and perspectives), Ground.news (aggregating news from left, center, and right sources to show multiple viewpoints), or university library research databases (providing access to journals from diverse publishers and viewpoints).

Complete access to all information sources globally. Government, corporate, academic, underground, alternative, competing viewpoints. Approaching god-like omniscient sourcing.

Real-World Example: No real-world example exists. Level ∞ would require access to every information source globally—government intelligence archives, corporate trade secrets, academic research, underground networks, dark web, competing ideological perspectives, and classified information from all nations—an impossible combination that approaches divine omniscience.
Return to Agent Power Framework