Table 2. Robo-investment aversion across studies.

Study	N	Design	Dependent variable	Effect size < 0 indicates robo-investment aversion
				Controversial stocks	Non-controversial stocks	Pooled
1	466	2 (between-subjects: human vs robo) × 2 (between-subjects: penalty vs no penalty)	Permissibility to exclude (Studies 1–3) or invest more heavily (Study 3) in stocks 1 = strongly disagree 5 = strongly agree	–0.25 [–0.44, –0.07]	–	–0.25 [–0.44, –0.07]
2	1,231	2 (between-subjects: human vs robo) × 2 (between-subjects: controversial vs non-controversial)		–0.47 [–0.62, –0.31]	–0.74 [–0.91, –0.57]	–0.58 [–0.69, –0.47]
3	683	2 (between-subjects: human vs robo) × 2 (within-subjects: exclusion vs inclusion)		–0.81 [–0.92, –0.70]	–	–0.81 [–0.92, –0.70]
4	705	2-cell (between-subjects: controversial vs non-controversial)	Choice of investment manager = 1 robo = 0 human	–0.09 [–0.30, 0.12]	0.26 [0.05, 0.47]	0.08 [–0.07, 0.23]
5	743	2-cell (between-subjects: controversial vs non-controversial)		–0.31 [–0.51, –0.11]	–0.28 [–0.49, –0.07]	–0.30 [–0.44, –0.15]
1–5	3,828					Mean effect fixed effects model –0.39 [–0.45, –0.32]
						Mean effect random effects model –0.36 [–0.64, –0.08]

Mean effect sizes are Cohen’s ds, with 95% confidence intervals reported in square brackets. Computed using the R package metaviz, based on the metafor package [58, 59]. Random effects are computed using the REML method.