PDF Publication Title:
Text from PDF Page: 031
The Problem in Focus: Factors and Remedies 21 Confounding Covariates Algorithm designers often choose to remove sensitive variables from their training data in an attempt to render the resulting system bias free (Barocas and Selbst, 2016). A common refrain by system design- ers is that “the system cannot be biased because it does not take [some sensitive variable] into account.” Barocas and Selbst (2016) discuss why just hiding sensitive variables often does not solve the problem. Machine learning methods can often probabilistically infer hidden variables, e.g., using ZIP Codes to infer (as a probabilistic substitute for) income. This ability also has strong implications for data privacy and anonymity. Researchers are reporting that traditional expecta- tions of data privacy and anonymity are no longer attainable (Dwork, 2008a; Narayanan and Shmatikov, 2010; Ohm, 2010). This is because modern machine learning algorithms are capable of reidentifying data easily and robustly. Remedies Remedying or regulating artificial agents will most likely require a combination of technical and nontechnical approaches. There are recent efforts underway to develop fair, accountable, and transparent machine learning techniques. We propose augmenting these with less technical approaches. Statistical and Algorithmic Approaches There is a growing field focused on fair, accountable, and transpar- ent machine learning, working on technical approaches to assuring algorithmic fairness or certifying and correcting disparate impact in machine learning algorithms. Dwork et al. (2012) proposed using modified distance or similarity metrics when working with subject data. These similarity metrics are meant to enforce rigorous fair- ness constraints when comparing subjects in data sets. Sandvig et al. (2014) proposed a number of algorithms auditing procedures that accounts with seemingly false names. The process for distinguishing between real and false names relied heavily on traditional Western naming practices (Hardt, 2014; Boyd, 2012).PDF Image | Intelligence in Our Image Risks of Bias and Errors in Artificial Intelligence
PDF Search Title:
Intelligence in Our Image Risks of Bias and Errors in Artificial IntelligenceOriginal File Name Searched:
RAND_RR1744.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
| CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com | RSS | AMP |