Remember to respond to two peers while being respectful of and sensitive to their viewpoints. Consider advancing the discussion in the following ways:
· Post an article, video, or visual to reinforce a peer’s idea or challenge them to see their point from a different perspective.
· Engage in conversation with your peers around the concept of the limits of inferential test results based on the human influences on the data quality and the data interpretation. Consider asking a question or sharing your own personal experience.
To complete this assignment, review the Psychology Undergraduate Discussion Rubric . You will also need:
· Norms of Practice for Online Discussion
PEER 1RW
Hello Class! I hope every one has a blessed week!
· What did you find most interesting about the video?
As someone who struggles to understand math, data, and algorithms I truly believed that algorithms were based on scientific data that was objective and true. I always hear people on social media apps such as Instagram and TikTok talking about the algorithm causing their content to be shadow banned or lose views, but I never actually understood what an algorithm
was. After watching this video, the most interesting part to me was learning that algorithms are not objective, true, or scientific and that they are simply opinions embedded in code (O’Neil, 2017). It was also eye opening to me to learn that we use basic algorithms in our every day lives!
· Why is it important to maximize the degree to which samples and data used to train or build algorithms are representative of the population?
It is vital to ensure that when building an algorithm, you are representing the entire population in your data to minimize the possibility of bias and to gather the most accurate information. If you only include certain pieces of information or small parts of a population, you may miss out on valuable data. One example of a failed algorithm is the algorithm that Amazon used for hiring people. Prior to 2015, Amazon used an algorithm for hiring new employees that was bias against women. The algorithm took the data from applications received over the course of 10 years to filter out the best candidates. This sounds great, but the problem was most of the applicants were males leading the algorithm to favor men over women.
· How can human bias influence data used to test hypotheses?
Every single person is bias in some form. Human bias can negatively influence data when used to test hypotheses by excluding parts of a population and resulting in flawed and unreliable conclusions. One thing that Cathy O’Neil mentioned that really resonated with me is how two people can have the exact same qualifications, but the person with a white-sounding name will get the job (O’Neil, 2017). As a society we need to do better and challenge all companies to do better and use representative samples when building algorithms.
· How does the potential for human influences on data selection and interpretation relate to one of the programmatic themes below?
While I believe that the potential for human influences on data selection and interpretation relate to both ethics and social justice, if I had to pick one over the other, I would say it better relates to social justice. The algorithms that are being used by companies, law enforcement, and even in the justice system are full of bias and are preventing every person from having human rights and access to the same opportunities.
References:
O’Neil, C. (2017, April). The Era of Blind Faith in Big Data Must End. TED Talks. Retrieved May 16, 2022, from
https://www.ted.com/talks/cathy_o_neil_the_era_of_blind_faith_in_big_data_must_end/t ranscript.
-Robin
PEER 2-AA
I had never thought of my daily life in terms of algorithms and formulas. I loved the way Cathy O’Neil used the example of making dinner and the different definitions she and her kid had of a successful meal, and how that success factor would alter their meal formulas. For her, it was eating veggies, and for her youngest son, it was eating lots of Nutella. Of course, there exists a formula in which both parties are satisfied, but I doubt O’Neil is going to feed her kids Nutella coated broccoli for dinner!
It is crucial when looking at data to look at as many samples and points as possible to create a fair and accurate representation of a population. If only one sample is observed, there exists a greater likeliness of the information portrayed not being representative of the whole population in the study. More samples=better accuracy. Human bias can factor in to this as well, as if the samples are not randomly selected, bias can show in who represents the sample. For example, a younger researcher studying moral beliefs may only look at young members of the population due to relating to their stance instead of including a variety of ages. This creates skewed data if the researchers hypothesis is of the general public, not just younger individuals.
Looking at the programmatic theme of ethics is key when observing human influences on the selection of relevant data. We all have inherent biases towards certain groups and ideas, and these biases must be somehow eliminated during data collection. To do this, researchers often want to eliminate the bias through random selection and equal vetting of participants. Population parameters must be well and clearly defined so as to not exclude certain members, and to make sure to include all relevant parties.
Source: O’Neil, C. (2017, April). The Era of Blind Faith in Big Data Must End. TED Talks. Retrieved May 17, 2022, from https://www.ted.com/talks/cathy_o_neil_the_era_of_blind_faith_in_big_data_must_end/t ranscript.
Reply to Thread
