Amazon Utilizes AI to Combat AI-Generated Reviews

2023-08-18

Amazon recently launched AI-generated customer review highlights, which summarize the themes and sentiments of written reviews to help shoppers quickly assess whether a product suits their needs. This summarization tool, which has been tested since the beginning of this year, is currently only available to certain mobile users in the United States.

A new AI-powered feature provides product insights and allows easy access to reviews that emphasize specific product attributes, such as "usability" or "performance".

By leveraging AI to present review highlights and encourage authentic feedback, Amazon aims to make the shopping journey for customers more transparent and clear.

Essentially, this technology derives its functionality from Amazon's community guidelines, which serve as parameters for machine learning models that analyze multiple data points to detect risks and rely on expert investigators to analyze fraud detection techniques to prevent fake reviews. The analysis includes various data points such as account relationships, login activities, review history, and indicators of abnormal behavior.

"Our goal is to ensure that every review in the Amazon store is trustworthy and reflects the buyer's actual experience. Amazon welcomes genuine reviews, whether positive or negative, but strictly prohibits fake reviews that intentionally mislead buyers," said David Montague, Vice President of Amazon's Seller Risk.

He added, "We continue to innovate our proactive technology to detect signs of fake reviews and other abnormal behavior."

This aligns with the company's "Rekognition Content Moderation" system, which is used to moderate harmful images in product reviews. The system combines machine learning with human involvement in the review process, starting with automated image decisions accounting for 40% and gradually increasing. Some self-managed models have transitioned to the Amazon Rekognition Detect Moderation API to improve accuracy.

This migration simplifies the architecture, reduces workload, and costs. The accuracy of Rekognition Content Moderation reduces the need and expense of human review, bringing significant benefits to the moderation of product reviews.

Amazon is strategically integrating AI into its product offerings. The company is no longer emphasizing prominent AI chatbots or image tools, but focusing on enabling developers to build their own generative AI tools using its AWS cloud infrastructure.

Recently, Amazon partnered with Meta to launch LLaMa 2, which runs on AWS. Although Amazon does not disclose detailed information about its AI/ML models, the "Review Summarization" tool is likely based on Meta's LLaMa 70B model.

Earlier this year, Amazon CEO Andy Jassy stated that generative AI has a significant impact on the company's future. This is evident from the generative AI initiatives being undertaken by various business departments at Amazon.

More Questions

However, the latest announcement about using AI to combat fake reviews has raised concerns about potential biases in the summarization process. While AI can compress a large amount of information into summaries, there are concerns that Amazon's profit motives may influence how AI presents information.

This could lead to a preference for high-profit products and well-known brands, potentially disadvantaging small sellers with limited marketing budgets.

Legal Action

Since 1995, customer reviews on Amazon have been an integral part of its platform, so it is wise to continuously improve their usefulness using AI. However, with the emergence of agents promoting fake reviews, reviews have largely lost credibility. Reports indicate that up to 40% of reviews on the platform may be fake.

Amazon's recent legal action against suspected promoters of fake reviews further demonstrates its commitment to combating fake reviews.

Montague stated, "Another way we combat fake reviews is through legal action. We not only target the source of the problem but also send a clear message to the outside world that there is no place for abuse in our store, and we will hold fraudsters accountable."

The Federal Trade Commission in the United States has recently proposed a rule to prohibit deceptive online reviews, aiming to enhance credibility. The development of this rule began in 2019 and involves cases related to false advertising and fake reviews. The proposed regulations prohibit the sale or solicitation of fake reviews, including fabricated profiles, AI-generated content, and non-user reviews, and impose penalties for violations.

Other prohibited activities include posting positive/negative reviews for the purchase of any product, allowing leadership/affiliated companies to review without proper disclosure, operating review websites as "independent" of their own products, suppressing reviews through threats/intimidation, and selling fake engagement metrics such as followers and video views.

Room for Improvement

Amazon's efforts have yielded results, with the company reporting that these methods have prevented over 200 million potentially fake reviews in the past year. The retail platform acknowledges that effective collaboration with private entities, consumer groups, and governments is crucial in addressing this issue.

Despite Amazon's efforts, consumer groups still believe that more measures need to be taken to combat the widespread problem of fake reviews. While Amazon's use of AI and legal action against operators of fake reviews has made progress, consumer groups emphasize the need for stronger legislative measures and further collaboration to ensure a genuine and trustworthy online shopping experience.

RECENT AI NEWS

RECENT AI TOOLS