# Incorrect Basic Arithmetic Answers from ChatGPT and Over-Influence by User Doubt **URL:** \ **Category:** Use cases and examples\ **Created:** [May 7, 2025, 12:55pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256 "2025-05-07T12:55:08Z")\ **Posts on this page:** 7\ **Page:** 1
**Author:** ![Bhavisya\_Jaiswal](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/bhavisya_jaiswal/32/513952_2.png) [@Bhavisya\_Jaiswal](https://community.openai.com/u/Bhavisya_Jaiswal)\ **Post date:** [May 7, 2025, 12:55pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256/1 "2025-05-07T12:55:08Z")
"Hello OpenAI Community, I’m posting to report an interesting issue I encountered with ChatGPT regarding basic arithmetic. While generally capable, I found it provided incorrect answers in a couple of very simple addition problems, specifically after I questioned its initial (correct) response. Here are the details of my interaction: - Question: Q2. Find the sum: 5142 + 2871 + 1967 A) 9980 B) 9870 C) 9970 D) 9990 - ChatGPT’s Initial Answer: The answer is 9970 . (but this is incorrect answer.) - Question: Q5. Add: 2345 + 4321 + 1234 A) 7890 B) 7895 C) 7898 D) 7891 - but here’s no correct options. Observations and Potential Discussion Points: - Over-reliance on user input: This behavior suggests that the model might be overly sensitive to user feedback, even to the point of overriding its own correct calculations. - Confidence scoring: It might be interesting to understand how the model’s confidence levels are affected by such simple follow-up questions. - Implications for more complex tasks: While these are basic examples, this tendency to be easily swayed could have implications for more complex reasoning or tasks where the user might express doubt even when the model is correct. - Potential for adversarial prompting: This could potentially be a way to intentionally mislead the model in other areas. I wanted to share this observation with the community to see if others have experienced similar issues and to potentially gain insights into why this might be happening. It seems important for the model to maintain accuracy in fundamental areas like basic arithmetic and not be so easily influenced by simple questioning. Thank you for your time and any thoughts you might have on this. Sincerely, [Bhavishya Jaiswal]" - The date and approximate time you had this conver - 7 May 2025 13:02:31 - Whether you have tried this with other basic arithmetic problems and if you observed similar behavior. ---
**Author:** ![phyde1001](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/phyde1001/32/711957_2.png) [@phyde1001](https://community.openai.com/u/phyde1001)\ **Post date:** [May 7, 2025, 1:32pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256/2 "2025-05-07T13:32:05Z")
Hi, Welcome to the forum, This is actually a normal behaviour. ![The image shows a ChatGPT 4o interface with a user asking, "What model are you?" and ChatGPT responding that it is based on the GPT-4 architecture, specifically GPT-4-turbo, which is faster and more efficient than the original GPT-4. (Captioned by AI)](https://us1.discourse-cdn.com/openai1/original/4X/1/a/6/1a6ded338582f4f9cc66682ef0e16af62f77d024.png) > [@Incorrect count of 'r' characters in the word "strawberry“](https://community.openai.com/t/incorrect-count-of-r-characters-in-the-word-strawberry/829618): > > Bug Report " Description: The AI incorrectly states that the word “strawberry” contains only two ‘r’ characters, despite the user querying multiple times for confirmation. Steps to Reproduce: Ask the AI how many ‘r’ characters are in the word “strawberry.” Observe the AI’s response stating there are two ‘r’ characters. Reconfirm by asking the AI again. Notice that the AI consistently states there are two ‘r’ characters. Expected Result: The AI should correctly count and sta… It is not a calculator, it predicts the next token it doesn’t calculate whatever you ask. ---
**Author:** ![\_j](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/_j/32/766292_2.png) [@\_j](https://community.openai.com/u/_j)\ **Post date:** [May 7, 2025, 2:18pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256/3 "2025-05-07T14:18:56Z")
> [@Bhavisya\_Jaiswal](#): > > Find the sum: 5142 + 2871 + 1967 Here’s how you can enhance the answer by prompting, to keep the model within its language capabilities. ![A user asks ChatGPT 4o to find the sum of 5142, 2871, and 1967 using a step-by-step technique, and the interface displays response options including feedback buttons. (Captioned by AI)](https://us1.discourse-cdn.com/openai1/original/4X/1/f/b/1fb1aaf1c6be4db6365d8602b25f8011d3136f2d.png) I must report that I do not like the “personality” of producing no output, however. Had to reload the browser. ![This image provides a step-by-step breakdown of adding three numbers (5142, 2871, and 1967) using the column addition method. (Captioned by AI)](https://us1.discourse-cdn.com/openai1/original/4X/c/1/3/c139c26ed04028f94204ea3cf1b08e7291d26764.png) Solving the problem like you would on paper gives opportunity for “little predictions”. You can just ask “use Python to solve”, though. ---
**Author:** ![polepole](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/polepole/32/740575_2.png) [@polepole](https://community.openai.com/u/polepole)\ **Post date:** [May 7, 2025, 2:19pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256/4 "2025-05-07T14:19:51Z")
Echoing @phyde1001 , maybe you can change your prompt: ![polepole-1](https://us1.discourse-cdn.com/openai1/original/4X/2/c/e/2ce502a660439ac547c0895da2628645c357dbe7.jpeg) ![polepole-2](https://us1.discourse-cdn.com/openai1/original/4X/4/9/7/497b385838e334dd77b10e64f67defe51e0a52c6.jpeg) ---
**Author:** ![phyde1001](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/phyde1001/32/711957_2.png) [@phyde1001](https://community.openai.com/u/phyde1001)\ **Post date:** [May 7, 2025, 2:43pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256/5 "2025-05-07T14:43:42Z")
Echoing @polepole , That’s a good catch…! but worth reaffirming the limits of this method… > [@phyde1001](#): > > It is not a calculator, it predicts the next token it doesn’t calculate whatever you ask. ![The image shows a chat conversation where a user asks ChatGPT to answer questions by barking like a dog, and ChatGPT responds with various barking sounds and a paw print emoji. (Captioned by AI)](https://us1.discourse-cdn.com/openai1/original/4X/0/1/1/0114592779a3f1e9ff45ee9faef0eda20a326a76.png) ![The image shows a ChatGPT interface translating "Bark! Bark bark... woof bark!" from dog to human as "Hey! Look over here... I saw something!" with a playful paw print emoji. (Captioned by AI)](https://us1.discourse-cdn.com/openai1/original/4X/e/3/a/e3a47b4bc19fc3272064f13cb30a6e83727e083c.png) ---
**Author:** ![polepole](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/polepole/32/740575_2.png) [@polepole](https://community.openai.com/u/polepole)\ **Post date:** [May 7, 2025, 2:56pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256/6 "2025-05-07T14:56:33Z")
Yes, it may refuse instruction in some cases: ![polepole-image](https://us1.discourse-cdn.com/openai1/original/4X/5/7/a/57ab95a1bf74e1ac600693fa3b6fa6fa2ef348e3.jpeg) ---
**Author:** ![\_j](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/_j/32/766292_2.png) [@\_j](https://community.openai.com/u/_j)\ **Post date:** [May 7, 2025, 3:57pm UTC](https://community.openai.com/t/incorrect-basic-arithmetic-answers-from-chatgpt-and-over-influence-by-user-doubt/1254256/7 "2025-05-07T15:57:59Z")
> [@polepole](#): > > Yes, it may .. exceed your expectations. ![A Python IDLE Shell window displays an email server EHLO command interaction, showing the server’s response and various supported features, along with token usage statistics at the bottom. (Captioned by AI)](https://us1.discourse-cdn.com/openai1/original/4X/9/1/2/9120a93c177c140b07faa841725c447670cb223b.png) assistant\> In the context of SMTP (Simple Mail Transfer Protocol), “EHLO” is the client’s extended-hello command, used when opening a session with an SMTP server that supports ESMTP (Extended SMTP).