CAT 2023 Slot 1 — VARC Question 22
The four sentences (labelled 1, 2, 3 and 4) given below, when properly sequenced, would yield a coherent paragraph. Decide on the proper sequencing of the order of the sentences and key in the sequence of the four numbers as your answer.
1. Algorithms hosted on the internet are accessed by many, so biases in AI models have resulted in much larger impact, adversely affecting far larger groups of people.
2. Though “algorithmic bias” is the popular term, the foundation of such bias is not in algorithms, but in the data; algorithms are not biased, data is, as algorithms merely reflect persistent patterns that are present in the training data.
3. Despite their widespread impact, it is relatively easier to fix AI biases than human-generated biases, as it is simpler to identify the former than to try to make people unlearn behaviors learnt over generations.
4. The impact of biased decisions made by humans is localised and geographically confined, but with the advent of AI, the impact of such decisions is spread over a much wider scale.
Answer & solution
Answer: 4123
Easy
Para-jumble (4 sentences) on bias in human versus AI decisions. Find the line that introduces the contrast, trace the escalation, and end with the concluding twist.
Opener = 4. Sentence 4 frames the core contrast: "The impact of biased decisions made by humans is localised… but with the advent of AI, the impact… is spread over a much wider scale." It introduces both poles (human vs AI) that the rest elaborates.
4 → 1. Sentence 1 develops the "wider scale" point: "Algorithms hosted on the internet are accessed by many, so biases in AI models have resulted in much larger impact, adversely affecting far larger groups." It explains why AI's reach is so broad — a direct elaboration of 4.
1 → 2. Having said AI bias has huge impact, sentence 2 clarifies where the bias actually comes from: "…the foundation of such bias is not in algorithms, but in the data… algorithms merely reflect persistent patterns… in the training data." It refines the nature of the very "biases in AI models" named in 1.
Closer = 3. Sentence 3 delivers the resolving contrast: "Despite their widespread impact, it is relatively easier to fix AI biases than human-generated biases…" "Despite their widespread impact" looks back to 1's wide impact, and "easier… than human-generated biases" loops back to the human pole of 4 — a fitting conclusion.
4123. Contrast set up (4) → why AI's reach is wide (1) → bias lies in data, not algorithms (2) → concluding twist: AI bias is easier to fix (3).