Future Directions of Fairness-Aware Data Mining: Recommendation, Causality, and Theoretical Aspects

Future directions of
Fairness-aware Data Mining
Recommendation, Causality, and Theoretical Aspects
Toshihiro Kamishima*1 and Kazuto Fukuchi*2
joint work with Shotaro Akaho*1, Hideki Asoh*1, and Jun Sakuma*2,3
*1National Institute of Advanced Industrial Science and Technology (AIST), Japan
*2University of Tsukuba, and *3JST CREST
Workshop on Fairness, Accountability, and Transparency in Machine Learning
In conjunction with the ICML 2015 @ Lille, France, Jul. 11, 2015
1

Outline
New Applications of Fairness-Aware Data Mining
Applications of FADM techniques, other than anti-discrimination,
especially in a recommendation context
New Directions of Fairness
Relations of existing formal fairness with causal inference and
information theory
Introducing an idea of a fair division problem and avoiding unfair
treatments
Generalization Bound in terms of Fairness
Theoretical aspects of fairness not on training data, but on test data
✤ We use the term “fairness-aware” instead of “discrimination-aware,” because the word “discrimination”
means classiﬁcation in a ML context, and this technique applicable to tasks other than avoiding
discriminative decisions
2
Foods for discussion about new directions of fairness in DM / ML

PART Ⅰ
Applications of
Fairness-Aware Data Mining
3

Fairness-Aware Data Mining
4
Fairness-aware Data mining (FADM)
data analysis taking into account potential issues of fairness
Unfairness Prevention
S : sensitive feature: representing information that is wanted not to
inﬂuence outcomes
Other factors: Y: target variable, X: non-sensitive feature
Learning a statistical model from potentially unfair data sets so
that the sensitive feature does not inﬂuence the model’s outcomes
Two major tasks of FADM
Unfairness Detection: Finding unfair treatments in database
Unfairness Prevention: Building a model to provide fair outcomes
[Romei+ 2014]

Anti-Discrimination
5
[Sweeney 13]
obtaining socially and legally anti-discriminative outcomes
African descent names European descent names
Arrested?
Located:
Advertisements indicating arrest records were more frequently
displayed for names that are more popular among individuals of
African descent than those of European descent
sensitive feature = users’ socially sensitive demographic information
anti-discriminative outcomes

Unbiased Information
6
[Pariser 2011, TED Talk by Eli Pariser, http://www.ﬁlterbubble.com, Kamishima+ 13]
To ﬁt for Pariser’s preference, conservative people are eliminated from
his friend recommendation list in a social networking service
avoiding biased information that doesn’t meet a user’s intention
Filter Bubble: a concern that personalization technologies narrow
and bias the topics of information provided to people
sensitive feature = political conviction of a friend candidate
unbiased information in terms of candidates’ political conviction

In the recommendation on the retail store, the items sold by the site
owner are constantly ranked higher than those sold by tenants
Tenants will complain about this unfair treatment
Fair Trading
7
equal treatment of content providers
Online retail store
The site owner directly sells items
The site is rented to tenants, and the tenants also sells items
[Kamishima+ 12, Kamishima+ 13]
sensitive feature = a content provider of a candidate item
site owner and its tenants are equally treated in recommendation

Ignoring Uninteresting Information
8
[Gondek+ 04]
ignore information unwanted by a user
A simple clustering method finds two
clusters: one contains only faces, and the
other contains faces with shoulders
A data analyst considers this clustering is
useless and uninteresting
By ignoring this uninteresting information,
more meaningful female- and male-like
clusters could be obtained
non-redundant clustering : ﬁnd clusters that are as independent
from a given uninteresting partition as possible
clustering facial images
sensitive feature = uninteresting information
ignore the inﬂuence of uninteresting information

Part Ⅰ: Summary
9
A belief introduction of FADM and a unfairness prevention task
Learning a statistical model from potentially unfair data sets so that
the sensitive feature does not inﬂuence the model’s outcomes
FADM techniques are widely applicable
There are many FADM applications other than anti-discrimination,
such as providing unbiased information, fair trading, and ignoring
uninteresting information

PART Ⅱ
10

Discussion about formal deﬁnitions and treatments of fairness
in data mining and machine learning contexts
Related Topics of a Current Formal Fairness
connection between formal fairness and causal inference
interpretation in view of information theory
New Deﬁnitions of Formal Fairness
Why statistical independence can be used as fairness
Introducing an idea of a fair division problem
New Treatments of Formal Fairness
methods for avoiding unfair treatments instead of enhancing
fairness
PART Ⅱ: Outline
11

Causality
12
[Žliobaitė+ 11, Calders+ 13]
sensitive feature: S
gender
male / female
target variable: Y
acceptance
accept / not accept
Fair determination: the gender does not inﬂuence the acceptance
statistical independence: Y ?? S
An example of university admission in [Žliobaitė+ 11]
Unfairness Prevention task
optimization of accuracy under causality constraints

Information Theoretic Interpretation
13
statistical independence between S and Y implies
zero mutual information: I(S; Y) = 0
the degree of inﬂuence S to Y can be measured by I(S; Y)
Sensitive: S
Target: Y
H(Y )
H(S)
I(S; Y )
H(S | Y )
H(Y | S)
Information theoretical view of a fairness condition

Causality with Explainable Features
14
[Žliobaitė+ 11, Calders+ 13]
sensitive feature: S
gender
male / female
target variable: Y
acceptance
accept / not accept
Removing the pure inﬂuence of S to Y, excluding the eﬀect of E
conditional statistical independence:
explainable feature: E
(confounding feature)
program
medicine / computer
medicine → acceptance=low
computer → acceptance=high
female → medicine=high
male → computer=high
An example of fair determination
even if S and Y are not independent
Y ?? S | E

Information Theoretic Interpretation
15
Sensitive: S
Target: Y
Explainable: E H(E)
the degree of conditional independence between Y and S given E
conditional mutual information: I(S; Y | E)
We can exploit additional information I(S; Y; E) to obtain outcomes
I(S; Y ; E)
H(Y )
H(S)
H(S | Y, E)
H(Y | S, E)
H(E | S, Y )
I(S; E | Y )
I(Y ; E | S)
I(S; Y | E)

Why outcomes are assumed
as being fair?
16
Why outcomes are assumed as being fair,
if a sensitive feature does not inﬂuence the outcomes?
All parties agree with the use of this criterion,
may be because this is objective and reasonable
Is there any way for making an agreement?
To further examine new directions, we introduce a fair division problem
In this view, [Brendt+ 12]’s approach is
regarded as a way of making
agreements in a wisdom-of-crowds way.
The size and color of circles indicate the
size of samples and the risk of
discrimination, respectively

Fair Division
17
Alice and Bob want to divide this swiss-roll FAIRLY
Total length of this swiss-roll is 20cm
20cm
Alice and Bob get half each based on agreed common measure
This approach is adopted in current FADM techniques

Fair Division
17
Alice and Bob want to divide this swiss-roll FAIRLY
Alice and Bob get half each based on agreed common measure
This approach is adopted in current FADM techniques
10cm
10cm
divide the swiss-roll into 10cm each

Fair Division
18
Unfortunately, Alice and Bob don’t have a scale
Alice cut the swiss-roll exactly in halves based on her own feeling
envy-free division: Alice and Bob get a equal or larger piece
based on their own measure

Fair Division
18
Unfortunately, Alice and Bob don’t have a scale
envy-free division: Alice and Bob get a equal or larger piece
based on their own measure
Bob pick a larger piece based on his own feeling
Bob

Fair Division
19
Fairness in a fair division context
Envy-Free Division: Every party gets a equal or larger piece than
other parties’ pieces based on one’s own measure
Proportional Division: Every party gets an equal or larger piece than 1/n
based on one’s own measure; Envy-free division is proportional division
Exact Division: Every party gets a equal-sized piece
There are n parties
Every party i has one’s own measure mi(Pj) for each piece Pj
mi(Pj) = 1/n, 8i, j
mi(Pi) 1/n, 8i
mi(Pi) mi(Pj), 8i, j

Envy-Free in a FADM Context
20
Current FADM techniques adopt common agreed measure
Can we develop FADM techniques using an envy-free approach?
This technique can be applicable without agreements on fairness criterion
FADM under envy-free fairness
Maximize the utility of analysis, such as prediction accuracy,
under the envy-free fairness constraints
A Naïve method for Classification
Among n candidates k ones can be classified as positive
Among all nCk classifications, enumerate those satisfying envy-free
conditions based on parties’ own utility measures
ex. Fair classifiers with different sets of explainable features
Pick the classification whose accuracy is maximum
Open Problem: Can we develop a more efficient algorithm?

Fairness Guardian
21
Current fairness prevention methods are designed so as
to be fair
Example: Logistic Regression + Prejudice Remover [Kamishima+ 12]
The objective function is composed of
classiﬁcation loss and fairness constraint terms
P
D ln Pr[Y | X, S; ⇥] + 2 k⇥k2
2 + ⌘ I(Y ; S)
Fairness Guardian Approach
Unfairness is prevented by enhancing fairness of outcomes

Fair Is Not Unfair?
22
A reverse treatment of fairness:
not to be unfair
One possible formulation of a unfair classifier
Outcomes are determined ONLY by a sensitive feature
Pr[Y | S; ⇤
]
Ex. Your paper is rejected, just because you are not handsome
Penalty term to maximize the KL divergence between
a pre-trained unfair classifier and a target classifier
DKL[Pr[Y | S; ⇤
]k Pr[Y | X, S; ⇥]]

Unfairness Hater
23
Unfairness Hater Approach
Unfairness is prevented by avoiding unfair outcomes
This approach was almost useless for obtaining fair outcomes, but…
Better Optimization
The fairness-enhanced objective function tends to be non-convex;
thus, adding a unfairness hater may help for avoiding local minima
Avoiding Unfair Situation
There would be unfair situations that should be avoided;
Ex. Humans’ photos were mistakenly labeled as gorilla in auto-
tagging [Barr 2015]
There would be many choices between to be fair and not to be unfair
that should be examined

Part Ⅱ: Summary
Relation of fairness with causal inference and information theory
We review a current formal deﬁnition of fairness by relating it with
Rubin’s causal inference; and, its interpretation based on information
theory
New Directions of formal fairness without agreements
We showed the possibility of formal fairness that does not presume a
common criterion agreed between concerned parties
New Directions of treatment of fairness by avoiding unfairness
We discussed that FADM techniques for avoiding unfairness, instead
of enhancing fairness.
24

PART Ⅲ
Generalization Bound
in terms of Fairness
25

Part Ⅲ: Introduction
There are many technical problems to solve in a FADM literature,
because tools for excluding speciﬁc information has not been
developed actively.
Types of Sensitive Features
Non-binary sensitive feature
Analysis Techniques
Analysis methods other than classiﬁcation or regression
Optimization
Constraint terms make objective functions non-convex
Fairness measure
Interpretable to humans and having convenient properties
Learning Theory
Generalization ability in terms of fairness
26

Conclusion
Applications of Fairness-Aware Data Mining
Applications other than anti-discrimination: providing unbiased
information, fair trading, and excluding unwanted information
Relation of fairness with causal inference and information theory
Formal fairness introducing an idea of a fair division problem
Avoiding unfair treatment, instead of enhancing fairness
Generalization bound in terms of fairness
Generalization bound in terms of fairness based on f-divergence
Additional Information and codes
http://www.kamishima.net/fadm
Acknowledgments: This work is supported by MEXT/JSPS KAKENHI Grant Number 24500194, 24680015,
25540094, 25540094, and 15K00327
28

Bibliography I
A. Barr.
Google mistakenly tags black people as ‘gorillas,’ showing limits of algorithms.
The Wall Street Journal, 2015.
⟨http://on.wsj.com/1CaCNlb⟩.
B. Berendt and S. Preibusch.
Exploring discrimination: A user-centric evaluation of discrimination-aware data mining.
In Proc. of the IEEE Int’l Workshop on Discrimination and Privacy-Aware Data Mining,
pages 344–351, 2012.
T. Calders, A. Karim, F. Kamiran, W. Ali, and X. Zhang.
Controlling attribute eﬀect in linear regression.
In Proc. of the 13th IEEE Int’l Conf. on Data Mining, pages 71–80, 2013.
T. Calders and S. Verwer.
Three naive Bayes approaches for discrimination-free classiﬁcation.
Data Mining and Knowledge Discovery, 21:277–292, 2010.
C. Dwork, M. Hardt, T. Pitassi, O. Reingold, and R. Zemel.
Fairness through awareness.
In Proc. of the 3rd Innovations in Theoretical Computer Science Conf., pages 214–226,
2012.
1 / 4

Bibliography II
K. Fukuchi and J. Sakuma.
Fairness-aware learning with restriction of universal dependency using f-divergences.
arXiv:1104.3913 [cs.CC], 2015.
D. Gondek and T. Hofmann.
Non-redundant data clustering.
In Proc. of the 4th IEEE Int’l Conf. on Data Mining, pages 75–82, 2004.
T. Kamishima, S. Akaho, H. Asoh, and J. Sakuma.
Considerations on fairness-aware data mining.
In Proc. of the IEEE Int’l Workshop on Discrimination and Privacy-Aware Data Mining,
pages 378–385, 2012.
Enhancement of the neutrality in recommendation.
In Proc. of the 2nd Workshop on Human Decision Making in Recommender Systems,
pages 8–14, 2012.
Fairness-aware classiﬁer with prejudice remover regularizer.
In Proc. of the ECML PKDD 2012, Part II, pages 35–50, 2012.
[LNCS 7524].
2 / 4

Bibliography III
Eﬃciency improvement of neutrality-enhanced recommendation.
In Proc. of the 3rd Workshop on Human Decision Making in Recommender Systems,
pages 1–8, 2013.
E. Pariser.
The ﬁlter bubble.
⟨http://www.thefilterbubble.com/⟩.
E. Pariser.
The Filter Bubble: What The Internet Is Hiding From You.
Viking, 2011.
A. Romei and S. Ruggieri.
A multidisciplinary survey on discrimination analysis.
The Knowledge Engineering Review, 29(5):582–638, 2014.
L. Sweeney.
Discrimination in online ad delivery.
Communications of the ACM, 56(5):44–54, 2013.
I. ˇZliobait˙e, F. Kamiran, and T. Calders.
Handling conditional discrimination.
In Proc. of the 11th IEEE Int’l Conf. on Data Mining, 2011.
3 / 4

Bibliography IV
R. Zemel, Y. Wu, K. Swersky, T. Pitassi, and C. Dwork.
Learning fair representations.
In Proc. of the 30th Int’l Conf. on Machine Learning, 2013.
4 / 4

Future Directions of Fairness-Aware Data Mining: Recommendation, Causality, and Theoretical Aspects

More Related Content

What's hot (20)

Similar to Future Directions of Fairness-Aware Data Mining: Recommendation, Causality, and Theoretical Aspects (20)

More from Toshihiro Kamishima (12)

Recently uploaded (20)

Future Directions of Fairness-Aware Data Mining: Recommendation, Causality, and Theoretical Aspects