The online giant Amazon employs artificial intelligence (AI) to generate summaries of product reviews. This is intended to expedite the purchasing process for customers who are under time constraints and do not want to sift through numerous consumer reviews.
The advancing development of artificial intelligence has led to a variety of applications that can assist us in various aspects of daily life. This includes E-commerce, where Amazon has developed an AI that creates automated product reviews. However, a recent investigation has revealed that this AI does not reliably fulfill its task, and the disadvantages outweigh the benefits for customers.
The Product Review Summary Function
In 2023, Amazon introduced an AI designed to automatically summarize product reviews, thereby streamlining the purchasing process for customers. The goal of this function is to extract the key points from numerous reviews and highlight the most important aspects of a product in a concise review. Customers are supposed to receive a compact overview, eliminating the need to laboriously read through hundreds of reviews and assisting them in their purchase decision.
Weaknesses in Product Evaluation
However, Bloomberg's investigation reveals that Amazon's AI does not reliably fulfill its task. Analyzing dozens of products, it was found that the AI is unable to appropriately filter and consider the highlights of the reviews. Negative aspects in the reviews are often overemphasized, and overall, positive and negative points of a product are not coherently listed. This leads to a distortion of the overall assessment and can mislead customers.
An example from the investigation involves canning jars. Despite averaging a rating of 4.5 stars, the AI highlighted rust stains on the jars as a negative aspect. However, this aspect was mentioned by only 16 users out of a total of 3000 reviews. Nevertheless, this aspect constituted one-third of the AI summary, resulting in a distorted representation of the product.
Missing or Incorrect Information in AI Summaries
In addition to overemphasizing negative aspects, the investigation also reveals that Amazon's AI sometimes fails to appropriately acknowledge important information in the reviews. For example, a backpack with over 21,000 reviews lacked the mention that inferior material was used for the shoulder straps. There were even instances where the AI turned a fitness device into a desk, rendering the product description inaccurate.
Customers have noticed these weaknesses in Amazon's AI, stating that the function negatively affects holiday sales. Despite this criticism, Amazon rejects the accusations and emphasizes that the AI helps customers find the right products, leading to increased sales. Additionally, the company continuously works on improving the system.
The Challenges in Developing AI Systems
The problems highlighted by the Bloomberg investigation illustrate the current challenges in developing AI systems. Artificial intelligence is not capable of independent thinking but merely complements the most probable next word. This can lead to misinterpretations and distortions, as is the case with Amazon's AI.
AI-generated texts are unable to replace human thinking and judgment. They can only serve as tools to extract and structure information. When using such AI functions, customers should always remain critical and complement the information with their own research and assessments.