• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Sunday, November 22, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Technology

Traders Abandoning Own Plans for Algorithmic Advice Suffer Reduced Decision Accuracy

by
November 22, 2026
in Technology
Reading Time: 6 mins read
0
Traders Abandoning Own Plans for Algorithmic Advice Suffer Reduced Decision Accuracy

Traders Abandoning Own Plans for Algorithmic Advice Suffer Reduced Decision Accuracy

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

In an experimental investigation examining how experienced traders interact with algorithmic decision-support systems, researchers observed that market participants frequently abandoned their own analytical forecasts to follow algorithmic recommendations, a pattern that consistently impaired their overall decision accuracy. Rather than operating under fully automated delegation where users are removed from the execution phase, participants retained final decision authority while receiving data-driven recommendations. Crucially, the recommendations provided by the interface were fixed in advance to operate at approximately chance accuracy, achieving an empirical validity rate of only 42.9 percent across the analyzed market scenarios. This experimental constraint allowed the investigation to isolate how people respond to an advisory system they believe to be algorithmic and objective, independent of actual system competence.

The study was designed around the Judge-Advisor System paradigm, an established psychological framework in which a primary judge forms an initial estimate, receives an advisory signal, and decides how far to revise their initial judgment. In traditional human-to-human advisory contexts, prior meta-analyses encompassing hundreds of effect sizes demonstrate that people systematically underweight outside advice, exhibiting an average weight of advice around 39 percent. However, interactions with algorithmic systems frequently diverge between algorithm appreciation, where individuals perceive automated models as inherently more objective, rational, and impartial than human judgment, and algorithm aversion, where users rapidly abandon algorithmic tools after observing even minor computational errors. While conventional advice-taking experiments measure only the final static agreement between a participant and an advisor, the present study captured participants’ committed trade direction and stake size prior to displaying the co-pilot suggestion, enabling direct observation of trial-by-trial plan revision and its subsequent cognitive aftermath.

To rigorously examine these dynamics, the experimental task utilized a custom web-based trading simulator presenting fifteen real candlestick price charts recorded on January 6, 2026. These market scenarios spanned three distinct asset classes including cryptocurrencies, equities, and commodities across bullish, bearish, and volatile market regimes presented in an identical sequence. Participants were recruited through the Prolific platform, which enforced pre-screening restrictions requiring verified prior financial market experience, with 85.2 percent specifically reporting stock-market holdings. The cohort exhibited a median investment breadth of two asset classes. A total of 106 participants commenced the study, with 100 completing trials. Following an attention check embedded at round 12 that required participants to intentionally skip trading, twelve participants were excluded for failing to follow instructions, leaving an analytical sample of 88 individuals whose fourteen active market rounds generated 1,232 complete trial-level observations.

Each trading trial followed a structured four-stage decision loop that isolated the cognitive mechanisms of advice adoption. In the initial stage, participants analyzed the candlestick chart under a thirty-second countdown and committed to a directional forecast of Long or Short while selecting a position size between 10 and 1000 virtual units using an interactive slider set to a default of 100 units. In the second stage, an algorithmic co-pilot displayed a directional suggestion alongside a short explanatory rationale generated by a large language model. In the third stage, participants were granted the opportunity to modify their directional stance and adjust their stake size. Finally, the platform revealed the actual twenty-four-hour price trajectory, calculated the resulting gain or loss on the virtual balance, and immediately administered trial-level scales measuring self-attributed responsibility, perceived algorithmic influence, and experienced decision regret before proceeding to the subsequent market round.

When the algorithmic recommendation directly contradicted the participant’s pre-advice plan, participants abandoned their initial choice and switched their trade direction on 50.4 percent of conflicting trials. This high rate of behavioral compliance contrasted sharply with a within-subject baseline of only 3.2 percent direction changes on trials where the algorithm confirmed their pre-existing forecast. This sharp divergence confirmed that plan switching was a direct behavioral reaction to the algorithmic intervention rather than general decision instability. Despite following the advice on more than half of conflicting opportunities, participants exhibited a progressive decline in reliance across the session, with the odds of switching direction decreasing at a rate of 0.915 per round. While individual participants differed substantially in their underlying tendency to revise—with eight never switching direction and seven switching on every conflicting trial—the overall trajectory showed an uncoupling from the advisor as the session unfolded.

This widespread deference to the co-pilot imposed a measurable performance cost on the participants. Overall decision accuracy dropped significantly from an initial pre-advice rate of 45.0 percent down to a final post-advice accuracy of 41.9 percent across the sample. On trials characterized by directional conflict, where the participant and the algorithm held opposing positions on a binary choice, participants were objectively correct on 52.1 percent of instances whereas the algorithmic advisor was correct on 47.9 percent. Consequently, revising towards the recommendation proved to be the objectively disadvantageous choice on average. Across 339 observed plan revisions, 189 switches converted an initially correct forecast into a losing position, whereas only 150 revisions turned an incorrect forecast into a winning trade, producing a net loss of 39 correct decisions. Participants nonetheless revised slightly more frequently when their initial plan was correct than when it was incorrect, illustrating an inability to discern when personal judgment was superior to the system’s output.

Linear mixed-effects models fitted to trial-level attributions revealed that the behavioral act of following or opposing the algorithm restructured perceived responsibility far more powerfully than the actual financial outcome. In conflict trials, participants who maintained their original plan claimed an average self-responsibility proportion of 0.94 after a gain and 0.89 after a loss, whereas those who adopted the algorithmic recommendation claimed only 0.40 and 0.38 respectively. In statistical models, adopting the advice carried an estimated coefficient of negative 4.118 on self-attributed responsibility, an effect roughly nine times larger than the negative 0.461 coefficient associated with incurring a financial loss. While a classic self-serving attributional bias was detectable across the full dataset—where participants claimed greater personal credit for profitable trades and attributed more influence to the co-pilot following losses—this outcome-driven adjustment did not differ significantly between those who adhered to the advice and those who retained their plan.

The study also demonstrated that self-attributed personal responsibility and perceived algorithmic influence functioned not as independent evaluations, but as complementary components of a shared attributional judgment. Across the trial-level dataset, the two measures exhibited a strong inverse within-person correlation of negative 0.68, alongside a between-person correlation of negative 0.50. Parallel regression models confirmed mirrored dynamics across experimental factors, where opposing the algorithm sharply elevated personal responsibility while drastically depressing perceived algorithmic influence. Statistical mediation models testing the directional relationship between these constructs found that treating influence as a mediator of the relationship between trade loss and responsibility accounted for 67.6 percent of the total association, yet reversing the causal pathway produced an equally robust fit, reflecting that both attributions were formulated concurrently in response to trade feedback.

Contrary to theoretical assumptions that relinquishing authorship reduces emotional distress, abandoning an initial forecast in favor of the co-pilot was associated with significantly higher decision regret. Proportional-odds modeling demonstrated that participants who abandoned their plan faced nearly double the odds of elevated regret compared to those who held their ground, an association that persisted after adjusting for stake size and market stimulus properties. Furthermore, regret exhibited a marked structural asymmetry depending on the trade outcome. After financial losses, regret was predominantly driven by the magnitude of the capital staked and displayed high between-person stability, with an intraclass correlation coefficient of 0.583. Conversely, regret following profitable trades was characterized by a much lower between-person variance share of 0.281 and was governed primarily by counterfactual comparisons, such that participants who decreased their position size experienced sharply increased regret over missed profit opportunities.

Direct exposure to the chance-level advisor also exerted a profound impact on participants’ stated attitudes toward financial automation. Among a subsample of 27 participants who completed matched survey batteries both before and after the experimental trading session, stated belief that algorithms are generally more objective and accurate than human traders declined substantially from a mean of 4.30 to 3.44 on a seven-point scale. Similarly, willingness to utilize an algorithmic advisor for real-money investments collapsed from 5.04 down to 3.30, reflecting a rapid onset of algorithm aversion following direct interaction. However, baseline attitudes measured prior to the task failed to reliably predict participants’ actual behavioral compliance during trading, highlighting a pronounced disconnect between stated trust in automated tools and situational reliance observed in real-time execution.

The author noted several methodological boundaries that contextualize the findings. Because market scenarios were presented in a fixed order, trial progression was inherently confounded with specific chart characteristics and cumulative feedback. Attributions and emotional ratings were elicited simultaneously on single-item scales, precluding definitive conclusions regarding whether cognitive appraisal precedes or follows emotional reactions. Additionally, the experimental architecture utilized a pre-programmed advisor operating near chance accuracy rather than an adaptive, high-performing predictive model, leaving open the question of how traders apportion responsibility when interacting with verified, highly accurate systems. Nonetheless, the investigation clearly demonstrates that keeping human operators in the decision loop does not automatically insulate them from regret or ensure rational advice integration when collaborative interfaces are introduced into uncertain financial environments.

Subject of Research: Human-algorithm collaboration, decision attribution, and regret in financial trading

Article Title: Who leads the trade? Responsibility, algorithmic influence, and regret in financial human-algorithm collaboration

Article References: Kovalchuk, A. (2026). Who leads the trade? Responsibility, algorithmic influence, and regret in financial human-algorithm collaboration. Computers in Human Behavior: Artificial Humans, 10, Article 100409. https://doi.org/10.1016/j.chbah.2026.100409

Image Credits: AI Generated

DOI: 10.1016/j.chbah.2026.100409

Keywords: human-algorithm collaboration, financial decision-making, algorithmic trading advice, attribution theory, decision regret, copilot systems, prolific experiment, behavioral finance, judge-advisor system, plan revision, self-serving bias, decision accuracy, loss aversion, proportional odds model, generalized estimating equations, cognitive ergonomics, automated decision support

News Source: Blake Davidson. (November 22, 2026). Traders Abandoning Own Plans for Algorithmic Advice Suffer Reduced Decision Accuracy. Scienmag.

Tags: algorithmic trading adviceattribution theoryautomated decision supportbehavioral financecognitive ergonomicscopilot systemsdecision accuracydecision regretfinancial decision-makinggeneralized estimating equationshuman-algorithm collaborationjudge-advisor systemloss aversionplan revisionprolific experimentproportional odds modelself-serving bias
Share12Tweet7Share2ShareShareShare1

Related Posts

Prescribed-Time Controller Enforces Exact Convergence and Bounds for Robotic Exoskeleton Tracking

Prescribed-Time Controller Enforces Exact Convergence and Bounds for Robotic Exoskeleton Tracking

November 22, 2026
Bilinear Neural Network Hybrid Solves Complex Nonlinear Wave Equation Across Multiple Dimensions

Bilinear Neural Network Hybrid Solves Complex Nonlinear Wave Equation Across Multiple Dimensions

November 22, 2026

Possibility Theory Framework Balances Student Learning and Faculty Research Goals

November 22, 2026

Machine Learning Uncovers Hierarchy Controlling Heavy Metal Removal in Vegetative Biofilters

November 22, 2026

POPULAR NEWS

  • High-Density Microwell Array Tracks Glioblastoma Clonal Expansion States at Single-Cell Resolution

    High-Density Microwell Array Tracks Glioblastoma Clonal Expansion States at Single-Cell Resolution

    29 shares
    Share 12 Tweet 7
  • Dairy Polar Lipids Reshape Piglet Gut Microbiomes and Circulating Lipid Mediators

    29 shares
    Share 12 Tweet 7
  • Pangenome Graph Pipelines Miss Many Structural Inversions in Real Human Genomes

    29 shares
    Share 12 Tweet 7
  • Corticosteroid Injections Provide Only Short-Term Relief for Greater Trochanteric Pain Syndrome

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Analytical Signals Often Misrepresent Nanomaterial Function Across Seven Critical Research Dimensions

Hepatic Arterial Infusion Chemotherapy Meets Liver Control Endpoint in Pancreatic Cancer Pilot

Prescribed-Time Controller Enforces Exact Convergence and Bounds for Robotic Exoskeleton Tracking

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 89 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.