5.3 - Lesson Notes

The hacks for this section are to summarize and give our takes on the discussion from class.

Algorithmic Bias

Computer bias often comes in the form of algorithms. Algorithms like to appeal to their common markets, such as Facebook including advertisements that are more appealing to the older generation while TikTok appeals to the youth.

We had a long discussion about female voices on A.I. systems. Ultimately, we decided that, while the roots of the decision may show signs of stereotyping, the fact that there are many more voices to choose from and that the creators likely collected consumer data about different voices makes it more likely that they picked a female default because it simply appealed more to the majority.

Algorithms are quite prominent in applications that want you glued to them: YouTube, Instagram, Netflix, Amazon, etc. While it can be harmful if, for example, someone consumes political content on YouTube and ends up stuck in an extremist echochamber without hearing other sides, usually these algorithms are to the benefit of the user. By curating content based on your past experience, sites like Netflix and YouTube get better at recommending media that you will enjoy most. I would argue that it is a net benefit for both parties.

Are HP Computers Racist?

No. While the data definitely doesn’t paint a very includive or responsible picture of HP testers, there isn’t any clear intent to discriminate behind the camera. Either way, it’s definitely a harmful quirk of the camera’s poor programming, and is definitely the fault of HP for not realizing.

I would argue that the best way to prevent this bad outcome is to test systems, especially those that are intended to work with people with lots of different appearances, with wide, diverse ranges of testers. A good way to collect a lot of data may be to film lots of different people in a public area.

5.4 - Lesson Notes

Crowdsourcing is using large groups of people to collect data. We can see it quite commonly with sites like Wikipedia, Swagbucks, and Opinion Outpost, and it has a hand in cryptocurrency, block chain algorithms, and COVID data.

Crowdsourcing exists in what we’ve learned about in this class. It can be seen with data in RapidAPI data sets and with Github. We should try to incorporate it into our group CPT projects.

Hacks

How can crowdsourcing be used for our CPT project?

  • CompSci has 150ish principles students. Describe a crowdsource idea and how you might initiate it in our environment.

It would be good to get feedback on our website’s games from other CSP students because it may help us get more detailed solutions or suggestions for various issues. This can apply to all branches of our project as well, and be initiated through Slack.

  • What about Del Norte crowdsourcing? Could your project be better with crowdsourcing?

Using a wider range of students would be ideal for testing the Reviews and Events databases, since their filtering and validation systems are better tested with many entries. If we got many Del Norte students to help make events and reviews on the site, we would have a better base to work with. However, this also increases the risk of spam.