Discuss about this article as it relates to Algorithms as one of the most potent forces affecting individual decision-making, the benefit of the algorithm and downside of it in atleast 50 words Algorithms have mastered checkers1, chess2,3, poker4, and tasks with fewer boundaries such as information search5. This expertise has led humans to rely heavily on algorithms. For example, people rely so heavily on Google that they treat it as an external memory source, resulting in them being less able to remember searchable information6. As big data has flourished, people have become so comfortable with algorithms that drivers will sleep in their self-driving cars7, go on dates with algorithmically-recommended matches8, and allow algorithms to run their retirement accounts9. However, there are some tasks for which humans prefer to take advice from other humans, such as in medical advice10 or predicting how funny a joke will be11. Humans often demonstrate greater reliance on advice from algorithms compared to non-algorithmic advice, exhibiting algorithmic appreciation12. Relying upon algorithms for analytical tasks is typically advantageous. Even simple algorithms, such as weighting all variables equally, can outperform human prediction13. In a meta-analysis of 136 studies, algorithms were 10% more accurate, on average, than non-algorithmic (human) judgment14. Consequently, for analytical tasks, we would expect a rational human to demonstrate algorithmic appreciation. Of course, much of human behavior is not strictly rational15. People tend to discount or disregard advice, even when it is not logical to do so16. Often, the source of advice dictates how much it is discounted. When people discount advice from other people less than they discount advice from algorithms, particularly after observing an algorithm make a mistake, they demonstrate algorithmic aversion—the opposite of algorithmic appreciation. There is evidence for both algorithmic aversion17 and appreciation12,18,19, and it is task dependent11. Prior research has also shown that people rely on advice more heavily when tasks become more difficult20. However, this effect may not be uniform across sources of advice. Given these empirical observations, we question whether task difficulty is an important explanatory variable in determining whether people demonstrate algorithmic appreciation or aversion. In our studies of reliance on algorithmic advice, we consider two critical factors: the source of advice and task difficulty. We conducted three preregistered experiments with N = 1500 participants to test the influence of algorithmic advice, compared to social influence, on human decision making. Broadly speaking, social influence encapsulates the myriad ways that humans change their behavior based on the actions of other people. Prior experiments show that when humans are exposed to social influence, the wisdom of the crowd can be reduced21, and that the structure of the social network dictates how social influence affects decision-making22. Based on subject responses across multiple tasks and under different manipulation conditions, we find that people rely more on algorithmic relative to social advice, measured using Weight on Advice (WOA)23. Further, we establish that advice acceptance varies as tasks increase in objective difficulty and as advice varies in quality.
Discuss about this article as it relates to Algorithms as one of the most potent forces affecting individual decision-making, the benefit of the
Algorithms have mastered checkers1, chess2,3, poker4, and tasks with fewer boundaries such as information search5. This expertise has led humans to rely heavily on algorithms. For example, people rely so heavily on Google that they treat it as an external memory source, resulting in them being less able to remember searchable information6. As big data has flourished, people have become so comfortable with algorithms that drivers will sleep in their self-driving cars7, go on dates with algorithmically-recommended matches8, and allow algorithms to run their retirement accounts9. However, there are some tasks for which humans prefer to take advice from other humans, such as in medical advice10 or predicting how funny a joke will be11.
Humans often demonstrate greater reliance on advice from algorithms compared to non-algorithmic advice, exhibiting algorithmic appreciation12. Relying upon algorithms for analytical tasks is typically advantageous. Even simple algorithms, such as weighting all variables equally, can outperform human prediction13. In a meta-analysis of 136 studies, algorithms were 10% more accurate, on average, than non-algorithmic (human) judgment14. Consequently, for analytical tasks, we would expect a rational human to demonstrate algorithmic appreciation.
Of course, much of human behavior is not strictly rational15. People tend to discount or disregard advice, even when it is not logical to do so16. Often, the source of advice dictates how much it is discounted. When people discount advice from other people less than they discount advice from algorithms, particularly after observing an algorithm make a mistake, they demonstrate algorithmic aversion—the opposite of algorithmic appreciation. There is evidence for both algorithmic aversion17 and appreciation12,18,19, and it is task dependent11. Prior research has also shown that people rely on advice more heavily when tasks become more difficult20. However, this effect may not be uniform across sources of advice.
Given these empirical observations, we question whether task difficulty is an important explanatory variable in determining whether people demonstrate algorithmic appreciation or aversion. In our studies of reliance on algorithmic advice, we consider two critical factors: the source of advice and task difficulty. We conducted three preregistered experiments with N = 1500 participants to test the influence of algorithmic advice, compared to social influence, on human decision making. Broadly speaking, social influence encapsulates the myriad ways that humans change their behavior based on the actions of other people. Prior experiments show that when humans are exposed to social influence, the wisdom of the crowd can be reduced21, and that the structure of the social network dictates how social influence affects decision-making22. Based on subject responses across multiple tasks and under different manipulation conditions, we find that people rely more on algorithmic relative to social advice, measured using Weight on Advice (WOA)23. Further, we establish that advice acceptance varies as tasks increase in objective difficulty and as advice varies in quality.
AI-generated content may present inaccurate or offensive content that does not represent bartleby’s views.
Unlock instant AI solutions
Tap the button
to generate a solution
