Who introduced vague sets?

Gau and Buehrer introduced the formulation in their 1993 paper.

Can a vague set represent an ordinary fuzzy set?

Yes, when every interval collapses to its corresponding point membership grade.

Is a vague set necessarily probabilistic?

No. Membership bounds do not acquire probability or confidence semantics without further assumptions.

AI Glossary / Graphs and symbolic AI / Advanced

Vague set

Definition

A vague set represents each element using separate degrees of evidence for membership and against membership. These determine an interval of possible membership grades rather than a single point. In the standard formulation, the two degrees lie between zero and one and their sum cannot exceed one.

What it means

A conventional fuzzy set assigns an element one membership grade, such as 0.7 for belonging to the category “suitable supplier.” A vague set retains two quantities: a truth-membership degree t and a false-membership degree f. The membership interval is [t,1-f]. The false-membership degree is therefore not the interval's upper endpoint.
The gap between the endpoints records membership uncertainty left unresolved by the representation. It is not automatically a statistical confidence interval or a probability that the supplier will perform well. The meaning of the numbers depends on how evidence is assessed and mapped into the model. A formally valid interval can still rest on poor evidence.

How it works

For every element x, specify t(x) and f(x) with 0≤t(x),f(x)≤1 and t(x)+f(x)≤1. Compute the upper endpoint as 1-f(x). The interval width is 1-t(x)-f(x).
When t+f=1, the interval collapses to a point, recovering an ordinary fuzzy membership value. That point can be intermediate, so zero width does not necessarily make membership crisp. Crisp membership takes the special values zero or one.
A decision or query procedure must specify how it uses the intervals. It might inspect lower bounds, upper bounds, widths, or a justified ranking rule. Choosing a midpoint is an additional modeling step that discards width information. Vague sets and standard intuitionistic fuzzy sets encode corresponding membership and nonmembership information; different presentations do not establish universal superiority of one formalism.

Three practical examples

A supplier-screening exercise records evidence supporting suitability, evidence opposing it, and unresolved assessments. The team preserves the unresolved portion rather than treating missing evidence as a negative assessment.
A document-search prototype represents uncertain membership in a category. Users can distinguish items with stronger supported membership from items that merely have a high possible upper bound, provided the retrieval rule is clearly defined.
A product-review analysis separates favorable, unfavorable, and undecided judgments under a stated coding procedure. A vague-set representation can retain that distinction, but its values do not automatically estimate future sales or the views of the whole customer population.

Think of it this way

Imagine a review board with three trays: support, opposition, and unresolved material. A vague value preserves room for the unresolved material to affect membership. Combining all three trays into one summary score may be convenient, but it conceals how much remains unsettled and why.

Worked example

Consider a fictional panel of ten equally weighted reviewers assessing whether a supplier is suitable. For this toy coding rule, favorable judgments contribute to t, unfavorable judgments to f, and undecided judgments to neither. Six favor Supplier A, two oppose it, and two remain undecided.
Thus t(A)=0.6 and f(A)=0.2. A's membership interval is [0.6,0.8], with width 0.2. The interpretation concerns this panel and coding rule; it is not a confidence interval for the percentage of all possible customers who would approve.
Supplier B receives four favorable, zero unfavorable, and six undecided judgments. Its interval is [0.4,1.0], with width 0.6. Both intervals have midpoint 0.7, yet B has substantially more unresolved membership. A rule requiring the lower bound to reach 0.55 would pass A and fail B. This is an illustrative rule, not a universal decision procedure.
Suppose A's two undecided reviewers later resolve their judgments, one favorably and one unfavorably. The totals become seven favorable and three unfavorable, giving [0.7,0.7]. The width is now zero, but membership remains the intermediate fuzzy grade 0.7, not crisp membership one.
An attempted assignment t=0.8 and f=0.4 would be invalid in this standard formulation because their sum is 1.2. Conflicting evidence needs an explicit handling policy; it cannot simply be entered as an admissible vague value.

Illustrative case study

In a fictional purchasing team, a dashboard originally displays only the midpoint of each supplier's assessment interval. Two candidates both score 0.7, so reviewers assume they have equivalent support.
After the interval endpoints and evidence notes are added, the team sees the distinction illustrated above: one candidate has a narrower range, while the other has many unresolved assessments. It requests additional information about the uncertain candidate instead of interpreting the high upper bound as strong endorsement.
The team also retains mandatory qualification checks outside this illustrative ranking method. It records reviewer weights, the treatment of abstentions, and how new evidence updates the values. The improved process comes from preserving information and clarifying interpretation, rather than assuming that the mathematical representation independently verifies supplier quality.

Common misconceptions and mistakes

“The upper membership bound is f.” It is 1-f.
“Membership and nonmembership must sum to one.” Their sum can be smaller, leaving an unresolved interval.
“Zero width means a crisp set.” It means a point membership value, which may lie strictly between zero and one.
“A high upper bound means strong support.” It may instead reflect limited evidence against membership and substantial unresolved information.
“Every pair of scores is valid.” The standard condition t+f≤1 must hold for each element.

Business use

Vague sets can be useful when an application needs to preserve supporting and opposing assessments separately while representing unresolved membership. Define the category, universe of candidates, evidence sources, and coding rules before calculating values.
Explain how the decision process uses an interval and what information its ranking rule loses. Test whether choices change under plausible alternative treatment of unresolved evidence. Keep the original assessments available for inspection, particularly when unequal expertise or duplicated evidence makes a simple reviewer-count scheme inappropriate. Compare the added complexity with the practical value of displaying the uncertainty; an interval is useful only when users understand what its endpoints represent.

Check your understanding

Answer: What interval follows from t=0.6 and f=0.2? [0.6,0.8].

Why can equal midpoints be misleading?

Answer: They hide differences in interval width and supported membership.

Why is t=0.8,f=0.4 inadmissible?

Answer: Their sum exceeds one, reversing the required ordering of the bounds.

Related terms

Membership function (mathematics)
Fuzzy set
membership function
intuitionistic fuzzy set
interval-valued membership
nonmembership degree
uncertainty representation
fuzzy query
evidence aggregation

Learn next

Compare point and interval membership representations. Then examine how different interval-ranking rules change a decision while preserving the same underlying assessments.

Sources

Lu and Ng, Vague Sets or Intuitionistic Fuzzy Sets for Handling Vague Data: Which One Is Better?: https://home.cse.ust.hk/faculty/wilfred/paper/er05.pdf Borumand Saeid, (alpha,beta,T)-Convex Vague Sets, introductory definitions: https://lematematiche.dmi.unict.it/index.php/lematematiche/article/download/897/823/3543

Previous: VACUUM Next: Validation