The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2485
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T07:41:07.443009Z
status: published
lens: g_space
sector: business
headline: Too little, too much, or just right? How AI recommendation information volume shapes trust and consumer behavior
dek: Artificial intelligence (AI) shopping assistants explain why a product is selected, yet the appropriate defaults remain unclear. We define recommendation information volume as the amount of nonredundant, product-relevant information and compare five ordered explanation packages. Study 1 identifies an inverted-U response in a randomized Chinese mobile-retail experiment. Study S1 compares identical messages attributed to an AI assistant or a human salesperson, with separate source-neutral measurement validation. S…
gain_title: In a series of studies including a randomized Chinese mobile-retail experiment and data on 51,248 recommendation impressions, moderate-sized AI explanation packages performed best among the tested sets.
problem_title: (none)
trace_subject: (none)
gain_reading: In a series of studies including a randomized Chinese mobile-retail experiment and data on 51,248 recommendation impressions, moderate-sized AI explanation packages performed best among the tested sets.
gain_evidence: Moderate packages perform best within the tested sets. | Study 1 identifies an inverted-U response in a randomized Chinese mobile-retail experiment.
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers tested how much explanation an AI shopping assistant should give when recommending a product. They compared five ordered explanation packages, from sparse to dense. A randomized Chinese mobile-retail experiment found an inverted-U response, and moderate packages performed best within the tested sets.

A behavioral study covered 51,248 recommendation impressions and service outcomes for 5000 distinct users, adding real-use evidence to the experiments. A further study compared identical messages attributed to an AI assistant or a human salesperson.

The work links benefits to perceived usefulness, trust in AI and decision comfort. Costs run through information overload and what the authors call algorithmic over-justification. The authors say the results support testing default explanations category by category.
limitation: Same-session estimates do not establish causal mediation, and because amount, diagnostic priority and content composition vary jointly, the experiments identify assigned-package effects, not pure quantity effects.
tag: Evidence-backed gain
key_points: Five ordered explanation packages were compared, defined by the amount of nonredundant, product-relevant information. | Study 1, a randomized Chinese mobile-retail experiment, found an inverted-U response to information volume. | Study 3 analysed 51,248 recommendation impressions and service outcomes for 5000 distinct users. | Study S1 compared identical messages attributed to an AI assistant or a human salesperson. | Positive indirect associations ran through perceived usefulness, trust in AI and decision comfort. | Negative indirect associations ran through information overload and algorithmic over-justification.
rundown: The research defines recommendation information volume as the amount of nonredundant, product-relevant information in an AI shopping assistant's explanation of why a product was chosen. Five ordered explanation packages were compared across several studies.

Study 1 was a randomized experiment in Chinese mobile retail and found an inverted-U pattern. Study 2 reported conditional positive indirect associations through perceived usefulness, trust in AI and decision comfort, and negative ones through information overload and algorithmic over-justification.

Study 3 supplied behavioral evidence from 51,248 recommendation impressions and service outcomes for 5000 distinct users. Moderate packages performed best within the tested sets. The authors explain this with bounded information processing and persuasion knowledge, across AI and human sources.
sources:
- peer_reviewed | Journal of Retailing and Consumer Services | https://doi.org/10.1016/j.jretconser.2026.105205 | 2026-11-07
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

7b01c96b7fb8361c7af9de8e46703ab281051268e6f447dd074862b8210432d6

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000