Distrust of Algorithms
Algorithm Aversion
Algorithm aversion is defined as a biased assessment of an algorithm, showing up as negative behaviour and attitudes toward it compared with a human agent. The operative comparison is in the definition: the same advice is rejected from a system and accepted from a person.
What keeps this out of the phobia vocabulary is that the opposite is equally documented. Algorithm appreciation names the situations where people trust and follow algorithmic advice over human recommendations. A tendency that reverses depending on the task is a discrimination being made, not a fear being felt, which is what separates it from technophobia.
Where the line falls
The pattern across domains is consistent enough to state as a rule. Algorithms are trusted for transactional work and distrusted for relational work.
In recruitment, algorithmic agents are trusted for salary negotiation and human recruiters are preferred for emotional support and career development, on the perception that a system cannot fulfil a relational role. In healthcare, patients resist AI diagnostics and treatment recommendations despite their proven accuracy, and the reason given is the perception that such systems lack empathy and cannot handle nuanced emotional interaction. Negative emotions are more likely to arise as the system's role in the decision grows.
Moral and emotional judgement is where the rejection is strongest. Algorithms are least trusted for ethical dilemmas and empathetic decisions, and people may reject algorithmic decisions specifically where the moral stakes are high, autonomous vehicle decisions and life-or-death medical situations being the named examples.
Why good news from a machine lands differently
One finding is sharper than the rest and has no obvious explanation in the material.
When a decision produces a positive outcome, consumers find it harder to internalise if an algorithm made it. When the outcome is negative, the reaction is much the same whether an algorithm or a person decided.
Good news is discounted when it comes from a machine and bad news is not. Whatever that reveals, it is not a uniform distrust: a person who genuinely disbelieved the system would discount both.
Why people prefer content labelled human
In marketing, AI influencers can be as effective as human ones at promoting products, and trust in their recommendations remains lower because consumers perceive human influencers as more authentic.
Participants also favour content explicitly identified as human-generated over AI-generated content, even when the quality of the AI content matches or exceeds it. That is a preference about provenance rather than about quality, and it is stated in the material as a finding rather than as a complaint.
How culture changes algorithm aversion
Individualistic cultures show a higher tendency to reject algorithmic recommendations, with the United States given as the example, attributed to an emphasis on autonomy and personalised decision-making. Collectivist cultures show lower aversion, with India given as the example, particularly where familiarity with algorithms is higher or where the decision aligns with social norms.
Familiarity is the variable worth noting there, because it is the one that changes over time and the one an organisation can act on. The others are dispositions.
Why this matters more than a preference
The stakes are set out plainly: algorithms have proven ability to outperform humans in many contexts, and resistance to their recommendations can lead to inefficiencies and suboptimal outcomes. The factors named as contributing are three:
- perceived accountability
- lack of transparency
- scepticism toward machine judgement, which is a narrower worry than machines turning against us
The first two of those are not irrational and are not properties of the person. An unaccountable and opaque system is one there are reasons to distrust, and calling the resulting caution an aversion locates the problem in the human. That framing is worth noticing on a term whose own literature describes the assessment as biased.
Told apart
Often confused with Distrust of Algorithms
- Fear of TechnologyTechnophobia is a general discomfort with devices and complexity. Algorithm aversion is specific, reverses by task, and is held by people entirely comfortable with technology, which is why the same person accepts an algorithm for a salary calculation and refuses one for a diagnosis.
- Fear of Robots Turning Against UsBoth concern machines making decisions and only one concerns the machine. Algorithm aversion is about the source of a recommendation and applies equally to a system with no physical form at all.
Common questions
Questions about Algorithm Aversion
- Is this a fear of machines?
- No, and the documented opposite is what rules it out. Algorithm appreciation describes the situations where people trust and follow algorithmic advice over human recommendations. A tendency that reverses depending on the task is a discrimination being made rather than a fear being felt.
- When are algorithms trusted?
- For transactional work. In recruitment, algorithmic agents are trusted for salary negotiation while human recruiters are preferred for emotional support and career development. The rejection is strongest for moral and emotional judgement, including autonomous vehicle decisions and life-or-death medical situations.
- Why do patients resist AI diagnosis?
- On the perception that such systems lack empathy and cannot handle nuanced emotional interaction, despite the proven accuracy of the diagnostics. Negative emotions are more likely to arise as the system's role in the decision grows, which means the resistance increases with the algorithm's involvement rather than with its error rate.
- What is the asymmetry in consumer reactions?
- Good news is discounted and bad news is not. When a decision produces a positive outcome, consumers find it harder to internalise if an algorithm made it, and when the outcome is negative the reaction is much the same either way. Someone who genuinely disbelieved the system would discount both.
- Does culture change it?
- Substantially. Individualistic cultures show higher rejection, with the United States as the example, attributed to an emphasis on autonomy and personalised decision-making. Collectivist cultures show lower aversion, with India as the example, particularly where familiarity with algorithms is higher or the decision aligns with social norms.
- Is the aversion irrational?
- Two of the three factors its own literature names are not. Perceived accountability and lack of transparency are properties of the system rather than of the person, and an unaccountable, opaque system is one there are reasons to be careful about. Only scepticism toward machine judgement is a disposition, and the term nonetheless describes the whole assessment as biased.
Added 2026-08-23 · Revised 2026-08-26
Nearby in the catalogue
Related entries
- Fear of TechnologyParliament made demolishing a machine a capital offence in 1727, and the Luddites, who arrived later, objected to something more specific than machines.
- Fear of Robots Turning Against UsAsimov coined it in 1947 and conceded in the same story that the fear had a certain justification, which is why the First Law exists at all.
- DoomscrollingDoomscrolling is reading bad news past the point where it helps. The word arrived in 2018, the behaviour was measured in the 1970s as mean world syndrome.
- Fear of Being Without a PhoneThe Oxford English Dictionary records nomophobia from 1803, meaning fear of laws. The phone version is a portmanteau produced by a survey in 2008.
- Fear of ChemicalsBy weight, 99.99 per cent of the pesticides in a diet are grown by the plants, and twenty-nine of the fifty-seven natural ones tested cause cancer in rodents.
- Fear of MachinesShredders, grinders, chainsaws and overlockers are the named triggers, and the usual route in is an accident. A 1931 book blamed sexual incompetence.