What causes race conditions between threads during parallel data processing?

Race conditions occur when multiple threads access and modify shared data simultaneously without proper synchronization. If one thread reads or writes data while another thread modifies it at the same time, the result can be unpredictable and lead to errors.

For example, if two threads increment a shared counter simultaneously without locks, the final value may be less than expected because read and write operations overlap.

To prevent race conditions, synchronization mechanisms such as locks, semaphores, mutexes, or atomic operations are used.