Supporting Studies

Disclaimer: This Manifesto is entirely independent of the studies presented and references their findings solely to inform its conceptual development.

Studies supporting “AI contributions must be transparent”

• Lessons learned: Transparency around using AI needs to be specific to maintain trust – Source
  Randomized experiment (N = 1,483) shows AI labels can lower organisational trust unless paired with precise source lists, highlighting nuance in disclosure.

• Labeling Messages as AI-Generated Does Not Reduce Their Persuasive Effects – PDF
  Survey experiment (N = 1,601) finds authorship labels boost transparency yet don’t affect persuasion.

Studies supporting “Collaboration is layered”

• When humans and AI work best together — and when each is better alone – Source
  Review of 106 experiments shows combinations excel when roles complement but underperform in algorithm-dominant tasks, underscoring role-layer matching.

• When combinations of humans and AI are useful: A systematic review and meta-analysis – Source
  Meta-analysis of 370 effect sizes finds content-creation tasks gain most from human-AI synergy, while decision tasks often suffer.

• Examining human–AI collaboration in hybrid intelligence learning environments – Source
  Classroom case study proposes a Synergy Degree Model; synergy fluctuates low-to-moderate and maps clear teacher/AI layers.

• Human-AI Co-Creativity: Exploring Synergies Across Levels of Creative Collaboration – PDF
  Framework details four collaboration levels (Digital Pen → AI Co-Creator) with empirical cases showing creative gains when responsibilities are explicitly layered.

• Human-generative AI collaboration enhances task performance but undermines human’s intrinsic motivation – Source
  Series of four studies (N = 3,562) finds immediate performance boosts yet drops in long-term motivation, stressing careful layering for sustainable teamwork.

Studies supporting “Responsibility stays with people”

• Influence of AI behavior on human moral decisions, agency, and responsibility – Source
  Drone-operator experiment shows AI advice shifts moral choices and reduces explicit responsibility, evidencing real “responsibility gaps.”

• AI Explainability: How to Avoid Rubber-Stamping Recommendations – Source
  Panel of experts and 1,221-respondent survey conclude explainability & human oversight are complementary safeguards; without explanations, oversight devolves to rubber-stamping.

• Accountable Artificial Intelligence: Holding Algorithms to Account – PDF
  Public-sector analysis urges transparent, interpretable models so humans can answer for algorithmic decisions (information-explanation-consequence triad).