A question’s position in the GMAT exam—how early or late you see a given question on the computer adaptive test— does not affect how much the question counts toward your score.
In general, the computer’s estimate of your ability changes more in response to a question appearing early in a section than to questions appearing later. That is because, in the beginning, the computer has fewer questions upon which to base its evaluation. Near the end of the section, this estimate is based on a greater number of questions and is, therefore, a more reliable estimate.
If you answer the first question correctly, for example, the computer can infer only that your ability is probably at or above the mean for the population taking the GMAT exam, but it cannot tell how far from the mean. It will then deliver a more difficult question, corresponding to a higher ability level, to begin the process of narrowing in on your true ability. If the computer delivers a question that is comparably difficult to the first question near the end of the section, it does so with the benefit of a considerable amount of information about your ability level. A correct response in this instance would not cause the computer to make as significant an adjustment to its estimate of your ability precisely because it is taking into account more than this single question in computing its estimate.