SCIENCE · INVESTIGATE AND EXPLAIN

Challenge equal training fits with new evidence

A model can match every available training point yet fail outside that range. Fit quality describes agreement with data used in construction; prediction quality requires new evidence and a plausible mechanism. Environmental forecasts also need uncertainty, boundary conditions and a domain where their assumptions remain defensible.

Learning outcome: Compare c 0,0.5 and 1 against the same new observation. Retain training and new-point errors separately; explain what additional data would distinguish a robust mechanism from a flexible interpolation.

Starting recap

Retain observation history, units, temporal coverage and model assumptions; distinguish a conditional reconstruction from detected or attributed environmental change.

All lessons are open directly. Your saved experiments remain when you leave and return.

Try

With c 0.5, both training errors are zero. At new x 3, linear prediction 3 and alternative 6 compared with observed 3 give squared errors 0 and 9. At x 4, predictions diverge further to 4 and 16.

Open the experiment

Predict

Identical training error guarantees:

Your prediction and reason

Experiment

Compare c 0,0.5 and 1 against the same new observation. Retain training and new-point errors separately; explain what additional data would distinguish a robust mechanism from a flexible interpolation.

Test equal training fits on a new evidence point

Worked teaching model · synthetic values · Original training data x 0,1,2 and y 0,1,2. Linear model y=x and alternative y=x+c x(x−1)(x−2) both fit exactly. Compare public new observation atx 3 and forecasts atx 4. One public test point cannot validate a general environmental forecast or serve as secure independent assessment.

Curvature coefficient: 0.5 model coefficient New observed response at x=3: 3 response units

Reset keeps your prediction, captures and practice history.

Original synthetic system; units differ by row
Quantity	Value	Unit
Linear training squared error	0	response²
Alternative training squared error	0	response²
Linear new-point prediction	3	response units
Alternative new-point prediction	6	response units
Linear new-point squared error	0	response²
Alternative new-point squared error	9	response²
Linear extrapolation at x=4	4	response units
Alternative extrapolation at x=4	16	response units

Sequence values (response units); not measured environmental data
Position	Supplied or candidate value	Linear candidate
x = 0	0	0
x = 1	1	1
x = 2	2	2
x = 3	6	3
x = 4	16	4

Identical training fit does not identify a unique extrapolation.

The alternative is deliberately constructed to match every training point. Public new-point performance can challenge it, but model choice, mechanism, independent future evidence and uncertainty remain necessary; reusing the test point for tuning changes its evidential role.

Predict, change one control, and save contrasting captures. Reconstruct the balance with units; then explain a process the ledger deliberately leaves out. Display precision is not measurement precision.

Identical training fit does not identify a unique extrapolation.

Observation to attach to a capture

Captures preserve recipes and replay their calculated results. They are synthetic model evidence, not physical measurements.

Understand

Define the evidence question

Reconstruct the observation or physical model

The original training points are(0,0),(1,1),(2,2). Both y=x and y=x+c x(x−1)(x−2) pass through all three because the extra term vanishes at each training x. The curvature coefficient can therefore change extrapolation without changing training error. More flexible form does not automatically mean better explanation.

Separate the competing interpretations

At x 3, the alternative predicts 3+6 c; at x 4 it predicts 4+24 c. Compare the squared error at a new supplied observation without refitting. If the new point is then used to choose or tune c, it has become part of model selection and should not continue to be described as an untouched test.

Challenge uncertainty and identification

This public example is a mathematical demonstration, not a climate projection or secure assessment. A single new observation may still be noisy or unrepresentative. A defensible evaluation uses independent evidence, uncertainty, competing mechanisms and explicit limits rather than claiming that one winning public case validates all future behavior.

Worked original packet

With c 0.5, both training errors are zero. At new x 3, linear prediction 3 and alternative 6 compared with observed 3 give squared errors 0 and 9. At x 4, predictions diverge further to 4 and 16.

Investigation

Changed-case transfer

At c 0.25 and new observed response 4, calculate both x 3 predictions/errors and x 4 predictions.

Feedback

Linear x 3 prediction 3 has squared error 1; alternative 4.5 has error 0.25. At x 4 they predict 4 and 10. Better performance on this one point does not prove the alternative is a general forecast.

Sources and evidence boundary

Original prose, numerical packets and exercises. Primary method/mechanism source, checked 30 October 2026. No current measured climate values, personal risk estimates, exposure limits or regulatory conclusions are inferred from these fictional cases. Retain first reasoning and critique in the course workspace. No physical pollutant exposure or environmental intervention is requested.

Apply

Public practice checks a specific claim. It does not establish independent mastery.

At c 0.25 and new observed response 4, calculate both x 3 predictions/errors and x 4 predictions.

Your explanation and limitations

Continue

Completion requires a saved prediction, at least one capture and a reflection. Presence checks do not grade your reasoning.

Skipping retains experiments and awards no mastery.

Previous: Use contrasting patterns without overstating an exact fit Next: Reproduce an analysis without overwriting its source records

Save a science backup or import one

Science work is stored locally in this browser, separately from mathematics and AI. It does not sync automatically. Use this science-only backup; this checkout's older all-subject export does not include science.

A legacy file is available only when the complete snapshot fits the 36 MB limit. For larger snapshots, prepare a complete file set below; never discard notebooks to make a smaller file.

Choose a science backup (maximum 36 MB)

Matching imported notebook drafts replace their current drafts. Omitted lessons and known first attempts remain. Export your current work first if you want to retain both drafts.

Complete file sets and course backups

A file set can hold more work than one science file. Keep its manifest and every part together. Each file is limited to 36 MB. This format needs this version of LearnAI or a compatible newer version; older versions can use only the legacy single-file backup above.

Export scope

The set includes every saved science course and global continuation at this moment. Later edits need a new export. A conflicting tab exports its own visible work, which may differ from saved work in another tab.

Choose the manifest and all parts together

0 file(s) selected. File order and filenames do not determine identity. Missing, changed or conflicting parts block the entire restore; no good part is applied by itself.

Science storage: loading. No account or remote computation is required.