Human & Artificial Intelligence: Navigating Cognitive Biases in Strategic Decision-Making

Topic
This Master Thesis examines why decision makers sometimes continue failing projects even when the facts suggest they should stop. This behavior is called escalation of commitment (EoC) and describes a cognitive bias in strategic decision-making. The study investigates whether advice from human advisors or artificial intelligence (AI) systems can reduce this bias. In an online business scenario, participants evaluated a failing expansion project and decided whether to continue or terminate it. The thesis also examines whether group pressure affects this decision.

Relevance
Strategic investment decisions are often made under uncertainty, prior commitments, and social pressure. In such contexts, managers, board members, project sponsors, and strategy or innovation leaders may continue failing projects because abandoning them can signal inconsistency, failure, or poor prior judgment. This thesis is relevant for investment, portfolio, and project-continuation decisions because it shows that human or AI advice does not automatically improve outcomes. Corrective advice may be insufficient once decision makers already hold a strong interpretation of the situation.

Results
The central finding is that decisions rarely changed once participants had made up their mind. Even when participants received advice from the beginning, this did not significantly reduce escalation of commitment. It also made no reliable difference whether the advice came from a human advisor or an AI system. Participants who first decided without advice rarely changed their decision after receiving advice later. Group conformity pressure also did not lead to more escalation. Instead, decisions remained stable.

Continuation was more often linked to future-oriented reasoning and skepticism toward advice, whereas termination was linked to financial evidence and opportunity costs.

Implications for practitioners

• Advice availability is not advice use.
  The study shows that simply making human or AI advice available does not mean that decision makers actually rely on it. Organizations should therefore focus on how advice is integrated into the decision process, not only on whether advice is provided.
• Do not rely on the advisor label alone.
  The study found no reliable difference between human and AI advice. For organizations, the key question is therefore not simply whether advice comes from a consultant or an AI system, but whether it is integrated into the decision process in a way that managers actually perceive their expertise.
• Look beyond the final vote.
  Continuation was often linked to future-oriented reasoning, former investments, and skepticism toward advice, while termination was linked to financial evidence, opportunity costs, and independent judgment. Decision makers should examine the reasoning behind strategic decisions, not only the decision outcome, to overcome cognitive biases.

Methods
This study used an online experiment with a business scenario. Participants took the role of a board member and evaluated a failing geographic expansion project. In total, 73 valid responses were analyzed. Some participants received advice from a human advisor or an AI system before their first decision. Others first decided without advice and received advice only afterwards. Later, all participants saw a group signal supporting continuation of the project. The study measured whether participants continued or terminated the project, how much budget they allocated, how confident they were, which reasons they gave, and which individual characteristics might be related to their decisions.