Social Explorer AI is designed to deliver the most accurate and reliable demographic data available. In some cases, a user request cannot be fulfilled exactly as stated due to limitations in data availability, methodology, geographic coverage, or dataset design.
Fallback rules define how the AI handles these situations consistently and transparently, ensuring users always receive the best available alternative.
Survey and Dataset Fallback
If the preferred dataset is unavailable for a requested year, table, or geography, the AI automatically applies a dataset fallback.
When possible, the AI prioritizes Social Explorer enhanced datasets. If those are unavailable, it falls back to the original source datasets.
If a requested survey does not support the selected geography, the AI switches to the closest valid survey that does.
Example
User request
“Show population by census tract in Alpine County, CA using ACS 1-Year.”
AI behavior
ACS 1-Year data is not available for census tracts in Alpine County due to population thresholds.
The AI switches to ACS 5-Year estimates, which support full tract coverage.
Variable and Table Fallback
If a specific variable or table is unavailable, discontinued, suppressed, or changed in definition, the AI selects the closest valid alternative.
This may include
using the most recent year where the variable exists
selecting a broader variable category
or substituting an equivalent table that preserves the original intent
Example
User request
“Give me detailed ancestry data for 2023.”
AI behavior
Detailed ancestry variables were discontinued after 2012.
The AI uses the most recent available ancestry data or broader ancestry categories where supported.
Geography Fallback
If a requested geography is unavailable, has changed boundaries, or is not supported by the dataset, the AI applies a geographic fallback.
This may include
switching to a higher-level geography
selecting a comparable boundary
or avoiding direct comparison when boundaries are inconsistent
Example
User request
“Compare poverty rates in Census Tract 123.45 from 2010 to 2020.”
AI behavior
Census tract boundaries changed between 2010 and 2020.
The AI explains that values are not directly comparable and recommends a consistent geographic level.
Comparison and Trend Fallback
If a requested comparison is methodologically invalid, the AI blocks the comparison and explains why.
Invalid comparisons include
mixing ACS 1-Year and ACS 5-Year estimates
comparing incompatible survey designs
or comparing years where definitions or universes differ significantly
Example
User request
“Compare median income using ACS 1-Year (2015) and ACS 5-Year (2020).”
AI behavior
The AI explains that comparing ACS 1-Year and 5-Year estimates is methodologically invalid and avoids producing misleading results.
Margins of Error and Small Sample Size Fallback
For small populations or rare demographic groups, estimates may be statistically unreliable.
In these cases, the AI warns users when margins of error are large or sample sizes are insufficient and avoids presenting misleading trend analysis.
Example
User request
“Show the change in Pacific Islander population in a small rural tract from 2015 to 2023.”
AI behavior
The AI explains that sample sizes are too small for reliable trend analysis and recommends using a larger geography.
Internal Knowledge Fallback
If no current survey data exists for a request, the AI may provide general demographic context based on internal knowledge.
This fallback is always clearly labeled and never presented as official survey data.
Example
User request
“What is the population trend in ZIP code 12345?”
AI behavior
No current survey data is available.
The AI provides general demographic context with a clear disclaimer that it is not based on current survey estimates.
Transparency and Documentation
Every fallback action is explicitly documented.
The AI always explains why a fallback occurred and which alternative was used.
All fallback outputs include
clear explanatory text
dataset and year attribution
Sources panel information
methodology notes where applicable
Why Fallback Rules Matter
Fallback rules prevent invalid or misleading analysis
They ensure users always receive the best available data
They maintain trust through transparency
They guide users toward valid alternatives rather than returning empty or incorrect results