Newegg, arguably the techies favorite e-commerce platform, has introduced a new way of summarizing customer reviews using artificial intelligence (AI), specifically using ChatGPT.
The purposes is to help shoppers make better-informed purchase decisions by offering summarized insights from customer reviews. This is a step beyond the average product ratings that customers might have relied on in the past.
Here’s how it works: ChatGPT analyzes all the customer reviews for a product and extracts key terms and phrases to create concise summaries called Review Bytes. These highlight the likes and dislikes of the product according to customer opinions. In addition to Review Bytes, there’s also a comprehensive paragraph called SummaryAI that delves deeper into the reviews, providing more in-depth insights.
So, for example, if someone is looking for a gaming monitor, they might find Review Bytes that cover positive aspects like response rate and resolution, and negative aspects like washed out blacks. The SummaryAI further enhances these insights.
These features are only available on Newegg’s desktop website, and only for products that have a sufficient number of reviews. The AI-generated content is clearly marked as such, allowing customers to understand how the summaries were created and to access the original reviews if they wish.
In theory, customers would be able to quickly grasp the key aspects of products without having to read through many reviews. As for manufacturers, it helps them understand what customers like and dislike, providing opportunities for product improvements.
How ChatGPT Works in General
Newegg would likely be utilizing ChatGPT’s advanced natural language processing capabilities, including tokenization, embedding, transformer-based processing, and summarization techniques, to analyze, interpret, and condense customer reviews into concise insights for shoppers.
ChatGPT processes text by breaking it down into chunks called tokens. In the context of customer reviews, this would involve breaking down each review into individual words or subwords, allowing the model to analyze them.
These tokens are then mapped to vectors in a continuous space using what’s known as embeddings. This represents the words in a numerical form that a neural network can process. A sequence of these vectors is used to capture the order and context of the words in the review.
ChatGPT relies on a transformer architecture, which uses self-attention mechanisms to understand the relationships between different parts of a text. In this case, it would allow the model to identify key phrases and sentiments in the reviews by understanding how different words relate to each other.
The model can then be trained or fine-tuned to summarize and extract specific insights from these encoded texts. For Newegg’s Review Bytes, this might involve identifying key phrases that represent the likes and dislikes about a product. For the SummaryAI, this might involve generating a more comprehensive summary that captures a broader range of customer opinions.
The original announcement from NewEgg mentions that customers can click on Review Bytes to see the full reviews from which the information was extracted. Although not explicitly mentioned, it’s also plausible that Newegg may be employing continuous learning techniques to adapt and improve the model over time, based on how customers interact with the summaries and what insights prove most useful.
Given the responsible labeling of AI-generated content, Newegg is likely also considering the ethical aspects of AI deployment, including being aware of and mitigating potential biases in the data and modeling process.
How to Game NewEgg’s AI Reviews
While AI-powered summarization tools like ChatGPT offer many advantages, they are not entirely immune to manipulation, especially if the underlying data (in this case, customer reviews) is skewed or tampered with. Here’s how someone with nefarious intentions might attempt to game the system to get favorable summaries for their product:
- Submitting Fake Reviews: By flooding the platform with positive fake reviews, a malicious actor could skew the overall sentiment. Since the AI relies on the data it analyzes, this influx of fake positivity could result in the summarizations being disproportionately favorable.
- Keyword Stuffing: If a person understands the key terms and phrases that the AI might be looking for (e.g., “easy to use,” “high quality”), they could craft reviews filled with these terms, even if they don’t genuinely reflect the product’s qualities. This could lead the AI to extract and highlight these phrases, thus presenting a skewed view.
- Multiple Accounts: Creating multiple accounts to post similar positive reviews could add a false sense of agreement and consensus among different users, leading the AI to consider these opinions as highly relevant.
- Exploiting Known Biases: If someone were to understand the specific biases or limitations within the AI’s algorithms, they might craft reviews that exploit these to create a more favorable impression.
- Negative Review Attacks on Competitors: Conversely, the attacker might flood a competitor’s product with negative reviews, using similar tactics to create an unfavorable impression, thus making their product look better in comparison.
- Collusion with Other Users: By coordinating with other users to post similar reviews or upvote specific opinions, a manipulator might artificially emphasize certain opinions, leading the AI to weigh those more heavily in the summarization.
- Adversarial Attacks: More sophisticated attackers might even craft reviews specifically designed to mislead the AI, using knowledge of the model’s architecture and training data. This could involve crafting text that appears legitimate to human readers but is interpreted differently by the machine learning model.
It’s important to note that platforms like Newegg are likely aware of these potential vulnerabilities and would typically implement various countermeasures. One approach is human moderation and automated filtering that can help identify and remove fake or suspicious reviews.
Machine learning can also be used to detect unusual patterns, such as multiple reviews from the same IP address or similar wording across different accounts. NewEgg, by allowing users to click through to the original reviews, will allow its customers to see the source of the summaries and judge for themselves if something seems off. However, regular monitoring and analysis of the review patterns and the AI-generated summaries will be needed to rectify any systematic biases or manipulations.
Even with these potential countermeasures, the risk of manipulation is a real challenge in any system that relies heavily on user-generated content, and it requires continuous vigilance and adaptation to combat effectively.
How Display Vendors Benefit from NewEgg’s Review Bites
Despite the many ways in which online reviews can be manipulated by bad actors, there are different ways the sellers and manufacturers of display products on Newegg can leverage the AI-powered review summarization to enhance their offerings, marketing, and engagement with customers. By understanding and applying these opportunities responsibly, they can create a more effective and customer-centric approach to selling on the platform. And what applies here will probably be relevant to most online shopping destinations as AI techniques like Review Bites are bound to become standard operational behavior.
| Opportunity | Description |
|---|---|
| Understanding Customer Feedback | Quickly grasp what customers like and dislike to guide improvements to products and marketing strategies. |
| Monitoring Trends | Reveal emerging trends or shifts in customer preferences to signal areas needing attention. |
| Tailoring Marketing Messages | Emphasize the most appreciated aspects of a product in marketing and advertising. |
| Enhancing Product Descriptions | Integrate insights from reviews into product descriptions, highlighting customer-praised aspects. |
| Responding to Negative Feedback | Address consistent negative points by improving the product or providing additional support. |
| Competitive Analysis | Understand how products compare to competitors, identifying both advantages and areas for improvement. |
| Improving Customer Engagement | Actively engage with and respond to reviews to foster community and show value for customer opinions. |
| Leveraging Transparency | Embrace transparency by encouraging customers to explore both summaries and original reviews. |
| Data-Driven Decision Making | Utilize aggregated insights from reviews for decisions about product development, inventory management, pricing strategies, etc. |
