Autonomy, Power and the Special Case of Scarcity: Consumer Adoption of Highly Autonomous Artificial Intelligence

Insignificant effects are observed regarding scarcity conditions, implying that external conditions (for example, time pressure and sense of urgency) influence the relationship between autonomy and its adoption.

“Insignificant effects are observed regarding scarcity conditions, implying that external conditions (for example, time pressure and sense of urgency) influence the relationship between autonomy and its adoption.”

Elevated AI autonomy engenders a sense of powerlessness that hinders adoption.

“[40,41] demonstrate that elevated AI autonomy engenders a sense of powerlessness that hinders adoption”

High autonomy leads to lower adoption intentions because it evokes feelings of powerlessness-a conflict between human and machine autonomy.

“Frank et al. [40], who propose that high autonomy leads to lower adoption intentions because it evokes feelings of powerlessness—a conflict between human and machine autonomy.”

As systems move toward truly agentic, transaction-executing commerce, the main limitation is measurement: a significant portion of the evidence deduces delegation from adoption or continuation intentions, rather than from direct observation of explicit delegation choices or verified execution.

“As systems move toward truly agentic, transaction-executing commerce [24,40,44,45,50,52,67], the main limitation is measurement: a significant portion of the evidence deduces delegation from adoption or continuation intentions, rather than from direct observation of explicit delegation choices or verified execution.”

Experimental autonomy manipulations demonstrate that increased autonomy can diminish acceptance via pathways of powerlessness.

“Experimental autonomy manipulations (e.g., auto-purchase vs. notify) demonstrate that increased autonomy can diminish acceptance via pathways of powerlessness [40,58].”

Task and contextual pressures influence autonomy responses: scarcity and time constraints can diminish the negative autonomy-to-powerlessness trajectory, rendering elevated autonomy more acceptable in urgent situations.

“Second, task and contextual pressures influence autonomy responses: scarcity and time constraints can diminish the negative autonomy-to-powerlessness trajectory, rendering elevated autonomy more acceptable in urgent situations [40,41].”

Design elements that give agents more freedom or make them feel more personal can make people less likely to buy.

“Some studies show that design elements that give agents more freedom or make them feel more personal can make people less likely to buy [23,40]”

Higher autonomy becomes more acceptable in high-urgency situations because speed and assistance become more valuable in comparison to autonomy concerns.

“This makes higher autonomy more acceptable because speed and assistance become more valuable in comparison to autonomy concerns [40,41].”

This helps explain why marketing outcomes change from one context to another.

“This helps explain why marketing outcomes change from one context to another (for example, [38,40,46]).”

Feelings of powerlessness and psychological reactance occur when cues of autonomy or personalization are perceived as threats to freedom or displacement, leading to decreased acceptance and heightened avoidance.

“First, feelings of powerlessness and psychological reactance occur when cues of autonomy or personalization are perceived as threats to freedom or displacement, leading to decreased acceptance and heightened avoidance [23,24,40].”

When systems signal greater autonomy, such as auto-execution, users may experience diminished control or agency due to feelings of powerlessness or perceived threats to freedom, thereby inhibiting adoption intentions.

“Several studies indicate that when systems signal greater autonomy (e.g., auto-execution), users may experience diminished control or agency due to feelings of powerlessness or perceived threats to freedom, thereby inhibiting adoption intentions [40].”

Within scarcity or urgency framing, users demonstrate a greater willingness to embrace higher autonomy, as the perceived value of speed and assistance escalates while the autonomy to powerlessness mechanism diminishes.

“The mapped evidence reveals a contextual override regime that mitigates the negative autonomy pathway: within scarcity or urgency framing, users demonstrate a greater willingness to embrace higher autonomy, as the perceived value of speed and assistance escalates while the autonomy → powerlessness mechanism diminishes [40,41].”

True agentic execution (placing an order and paying for it) is still rare and mostly limited to scenario manipulations rather than embedded transaction observation.

“In the same way, true agentic execution (placing an order and paying for it) is still rare and mostly limited to scenario manipulations rather than embedded transaction observation [40,41].”

In everyday shopping situations, executor-like behavior can stop people from adopting something by making them feel powerless and can make reactance worse when recommendations seem to be steering or limiting.

“In everyday shopping situations, executor-like behavior can stop people from adopting something by making them feel powerless [40] and can make reactance worse when recommendations seem to be steering or limiting [23].”

Changes that change the agent from advisor to executor (like auto-purchase vs. notify/assist) can lower acceptance through powerlessness pathways.

“Similarly, changes that change the agent from advisor to executor (like auto-purchase vs. notify/assist) can lower acceptance through powerlessness pathways [40].”

Marketing lift depends on maintaining the user's sense of control, not just on improving capability cues.

“This demonstrates that marketing lift depends on maintaining the user's sense of control, not just on improving capability cues (e.g., [22,23,40]).”

Negative reactions arise when autonomy cues suggest a loss of control (reactance/threat-to-freedom) or displacement (powerlessness); conversely, positive outcomes are observed when systems facilitate friction reduction while maintaining perceived authorship and recourse.

“Negative reactions arise when autonomy cues suggest a loss of control (reactance/threat-to-freedom) or displacement (powerlessness); conversely, positive outcomes are observed when systems facilitate friction reduction while maintaining perceived authorship and recourse [23,24,40].”

Particularly in regular, non-urgent shopping contexts, the user experience can shift toward helplessness and threat-to-freedom reactions, resulting in lower adoption intentions and higher avoidance.

“Particularly in regular, non-urgent shopping contexts, the user experience can shift in these situations toward helplessness and threat-to-freedom reactions, resulting in lower adoption intentions and higher avoidance [23,40].”

The autonomy-to-powerlessness mechanism is well-supported in controlled experiments but often not tested in real-life transactions.

“The autonomy-to-powerlessness mechanism is well-supported in controlled experiments, but it is often not tested in real-life transactions [40,46].”

Personalization can cause avoidance through threat-to-freedom mechanisms, and high autonomy can decrease adoption through helplessness in routine contexts.

“This directly addresses findings that personalization can cause avoidance through threat-to-freedom mechanisms [23] and that high autonomy can decrease adoption through helplessness in routine contexts [40].”

A lot of the strongest causal evidence in the corpus comes from experiments instead of natural settings, leaving the real-world persistence of effects under-identified.

“This suggestion fits with the fact that a lot of the strongest causal evidence in the corpus comes from experiments instead of natural settings, leaving the real-world persistence of effects under-identified [8,40].”

More autonomy can make people less accepting by making them feel powerless.

“These studies show that more autonomy can make people less accepting by making them feel powerless.”

Experiments confirm autonomy to powerlessness/reactance mechanisms and interface-driven reactance shifts, whereas surveys focus on usefulness/satisfaction/trust chains.

“This disparity is evident in both experimental and survey methodologies: experiments confirm autonomy → powerlessness/reactance mechanisms [23,40] and interface-driven reactance shifts [24], whereas surveys focus on usefulness/satisfaction/trust chains [42,44,69].”

Autonomy effects are contingent upon context especially regarding urgency and scarcity, and may not extend to routine shopping without education, safeguards, and trust recalibration.

“The mapped evidence suggests that autonomy effects are contingent upon contextespecially regarding urgency and scarcity-and may not extend to routine shopping without education, safeguards, and trust recalibration [40,41].”

Genuinely agentic execution (cart/checkout/payment) and negotiation are still uncommon, and the empirical base is still focused on low-autonomy, text-based assistants with outcomes dominated by intentions and satisfaction constructs.

“First, genuinely agentic execution (cart/checkout/payment) and negotiation are still uncommon, and the empirical base is still focused on low-autonomy, text-based assistants with outcomes dominated by intentions and satisfaction constructs [10,40].”

Autonomy should be seen as a boundary condition that affects the delegation mechanism rather than as a simple main effect.

“This suggests that autonomy should be seen as a boundary condition that affects the delegation mechanism rather than as a simple main effect [40,46].”

Experimental research frequently substantiates causal assertions when randomization is evident and manipulation checks are documented, but limitations endure, including scenario-based autonomy, restricted domains, and intention-laden outcomes.

“Table 11 demonstrates that experimental research frequently substantiates causal assertions when randomization is evident and manipulation checks are documented; nonetheless, limitations endure, characterized by scenario-based autonomy, restricted domains, and intention-laden outcomes [40,41,46].”

Negative pathways often come up through governance-related mediators, like powerlessness and reactance, which make adoption and delegation less likely even when benefits are clear.

“This is why negative pathways often come up through governance-related mediators, like powerlessness and reactance, which make adoption and delegation less likely even when the benefits are clear [23,40].”

Acceptance models usually assume that delegation is the action-level construct, and autonomy is the boundary condition that makes governance costs important.

“Acceptance models usually assume that delegation is the action-level construct, and autonomy is the boundary condition that makes governance costs important [23,24,40].”

Purchase intention/adoption intention is mostly positive via utility/trust, but can decrease under high autonomy.

“Mostly positive via utility/trust, but can decrease under high autonomy”

High-autonomy cues can reduce intention via powerlessness/threat-to-freedom, unless urgency/scarcity raises benefit salience and attenuates resistance.

“High-autonomy cues can reduce intention via powerless-raises benefit salience and attenuates resistance. ness/threat-to-freedom, unless urgency/scarcity”

Behavioral lift appears when systems reduce friction (latency/steps) or improve option fit (recommendation relevance).

“Behavioral lift appears when systems reduce fric-havioral intention rather than verified execution. tion (latency/steps) or improve option fit (recom-mendation relevance).”

The utility chain dominates: quality/usefulness leads to attitude/satisfaction, which leads to intention.

“Utility chain dominates: quality/usefulness → at-titude/satisfaction → intention.”

Observed conversion/purchase behavior is positive when friction drops or recommendations fit, but not universal.

“Positive when fric-tion drops or rec-ommendations fit, but not universal”

In many studies, conversion is proxied via self-reported ordering/behavioral intention rather than verified execution.

“In many studies,”