Step 2: The Contradiction Finder

Analyze all loaded papers and surface the most significant points where two or more authors make directly contradicting claims.

Rules

Only flag real contradictions — mutually exclusive claims about the same phenomenon
Exclude cases where differences are only in emphasis or scale
Aim for 5-10 contradictions. If fewer than 5 exist, list all and explain why there are few

Present results as a table:

Contested Claim	Position A (Paper, Year)	Position B (Paper, Year)	Root Cause of Disagreement
{claim}	{author, year}: {their position}	{author, year}: {their position}	{methodology / dataset / time period / definition / other}

For root cause of disagreement, choose from: , , , , or (with explanation).

Analyze all loaded papers and surface the most significant points where two or more authors make directly contradicting claims.

Only flag real contradictions — mutually exclusive claims about the same phenomenon
Exclude cases where differences are only in emphasis or scale
Aim for 5-10 contradictions. If fewer than 5 exist, list all and explain why there are few

Present results as a table:

Contested Claim	Position A (Paper, Year)	Position B (Paper, Year)	Root Cause of Disagreement
{claim}	{author, year}: {their position}	{author, year}: {their position}	{methodology / dataset / time period / definition / other}

For root cause of disagreement, choose from: , , , , or (with explanation).

Excuse	Reality
"The papers mostly agree"	Disagreements exist at methodology, scope, or definition level. Dig deeper.
"This is a nuance, not a contradiction"	If both claims cannot be simultaneously true, it's a contradiction.
"I need to summarize the papers first"	You already have the processing protocol. Work from the loaded papers directly.