Page 270 - Touhpad Ai

Bias awareness means understanding that AI systems can show unfair preferences due to factors like training data
used to train the AI models, rules they follow, the algorithms they use, or the principles with which the AI model was
designed. This awareness involves understanding that AI may occasionally make biased decisions because of how
AI model was developed or trained.

AI REBOOT

Answer the given questions:
1. Who comes to your mind when you think of a teacher or cook? Do you think you are biased?

2. What do you mean by bias awareness?

3. AI’s training data can have bias. (State True or False)

AI Bias

AI bias is a phenomenon that occurs when algorithm results are systematically
biased against a certain gender, language, race, wealth, etc. AI bias leads to
a skewed output. Algorithms can have inherent biases because AI models are
created by individuals with conscious or unconscious preferences, and these
preferences, which may not be discovered until the algorithm is used publicly
(as we saw in the case of the hand dispenser and Google’s photo tagging
algorithm).
Bias is one of the biggest challenges facing AI. Although all programmers try to have absolute factual data, there is an
inevitable bias when exploring the depth to which AI can be used. An inherent problem with AI systems is that they
are as good or as bad as the trained data. Bad data usually carries racial, gender, community, or ethnic biases in
algorithms responsible for critical decisions go unrecognised, they can lead to unethical and unfair consequences. In
the future, these biases may worsen, especially as many AI recruitment systems continue to rely on incorrect data for
training. Therefore, there is an urgent need to train these systems with unbiased data and to develop algorithms that
are easy to interpret.

Sources of Bias

Data bias happens when some parts of a dataset are given too much
weight or are over-represented implying the dataset doesn’t accurately
reflect what the machine learning model is meant to do, leading to unfair
outcomes and poor accuracy.
Often, biased outcomes often result in discriminate against certain groups
of people, such as those based on age, race, culture, or sexual orientation.
As AI systems become more common, the risk of data bias is that it can
amplify existing discrimination. To prevent any ambiguity or reduce the
biasness, it is important to identify the source of bias and take the necessary steps.
Addressing AI bias involves thorough examination of datasets, machine learning algorithms, and other elements of
AI systems to identify sources of potential bias.

268 Touchpad Artificial Intelligence - XI